[go: up one dir, main page]

CN119049041A - Litchi nondestructive sorting method and device, storage medium and computer equipment - Google Patents

Litchi nondestructive sorting method and device, storage medium and computer equipment Download PDF

Info

Publication number
CN119049041A
CN119049041A CN202411133719.2A CN202411133719A CN119049041A CN 119049041 A CN119049041 A CN 119049041A CN 202411133719 A CN202411133719 A CN 202411133719A CN 119049041 A CN119049041 A CN 119049041A
Authority
CN
China
Prior art keywords
image
litchi
features
prediction model
target image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202411133719.2A
Other languages
Chinese (zh)
Inventor
王林琳
方熙鹏
毛亮
李�杰
林少臻
胡峰贤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Vocational And Technical University
Original Assignee
Shenzhen Vocational And Technical University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Vocational And Technical University filed Critical Shenzhen Vocational And Technical University
Priority to CN202411133719.2A priority Critical patent/CN119049041A/en
Publication of CN119049041A publication Critical patent/CN119049041A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/41—Refractivity; Phase-affecting properties, e.g. optical path length
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/55—Specular reflectivity
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/34—Smoothing or thinning of the pattern; Morphological operations; Skeletonisation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/54—Extraction of image or video features relating to texture
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/56—Extraction of image or video features relating to colour
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/68—Food, e.g. fruit or vegetables

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Analytical Chemistry (AREA)
  • Chemical & Material Sciences (AREA)
  • Pathology (AREA)
  • Immunology (AREA)
  • Biochemistry (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • General Engineering & Computer Science (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Image Analysis (AREA)

Abstract

The application relates to a litchi nondestructive sorting method, a device, computer equipment and a storage medium based on multispectral imaging and deep learning, wherein the litchi nondestructive sorting method comprises the steps of obtaining multispectral images of litchi, wherein the multispectral images comprise at least two of visible light images, infrared light images and ultraviolet light images; the method comprises the steps of inputting the multispectral image into an image segmentation network for segmentation processing to obtain an effective image area of the litchi, extracting target image characteristics of the effective image area, inputting the target image characteristics into a trained prediction model, and outputting the actual grade of the litchi through the prediction model. The litchi quality classification is realized by a litchi quality prediction model based on multispectral imaging and deep learning, so that the problems of poor timeliness, high cost, strong subjectivity and the like in the prior art when the litchi quality is detected manually can be solved.

Description

Litchi nondestructive sorting method and device, storage medium and computer equipment
Technical Field
The application relates to the technical field of artificial intelligence, in particular to the technical field of nondestructive testing, and particularly relates to a nondestructive sorting method and device for litchis, computer equipment and a storage medium.
Background
Litchi is one of the original fruits in China, and has a long history and rich cultural connotation. The planting of litchi and related industries are dominant and special industries which are developed in Guangdong province, and litchi production plays an important role in the economic development of Guangdong province. In recent years, the yield of litchi is kept at a higher level, but with the continuous improvement of the living standard of people, the market demand for high-quality litchi is also growing.
However, if the quality of litchi cannot be accurately distinguished, the postpartum additional value of the litchi can be directly affected. The quality of litchi is often identified by manual visual inspection, and the manual inspection method cannot meet the requirements of modern agricultural development due to the defects of poor timeliness, high cost, strong subjectivity and the like. In order to meet market demands and promote the modernization process of the litchi industry, an efficient and accurate litchi quality assessment method is urgently needed.
Disclosure of Invention
Based on the problems, the application mainly aims to provide a litchi nondestructive sorting method, a device, computer equipment and a storage medium based on multispectral imaging and deep learning, so as to solve the problems of poor timeliness, high cost, strong subjectivity and the like in the prior art when the litchi quality is detected manually.
In order to achieve the aim, the first aspect of the application provides a litchi nondestructive sorting method, which comprises the steps of obtaining a multispectral image of litchi, inputting the multispectral image into an image segmentation network to be subjected to segmentation processing to obtain an effective image area of the litchi, extracting target image characteristics of the effective image area, inputting the target image characteristics into a trained prediction model, and outputting the actual grade of the litchi through the prediction model.
Optionally, the extracting the target image features of the effective image area comprises extracting initial image features of the effective image area, and screening target image features from the initial image features, wherein the number of the initial image features is larger than that of the target image features.
Optionally, the multi-spectral image is input into an image segmentation network for segmentation processing to obtain an effective image area of the litchi, and the method comprises the steps of converting the multi-spectral image into a gray image, performing binarization processing on the gray image to obtain a binarized image, performing morphological operation on the binarized image to obtain a morphological image, wherein the morphological operation comprises at least one of expansion, corrosion, open operation and close operation, and extracting edges of the morphological image through an edge detection algorithm to obtain the effective image area of the litchi.
Optionally, the extracting the target image features of the effective image area comprises performing RGB channel separation on the effective image area, respectively extracting RGB image data of red, green and blue channels, performing statistical calculation on the RGB image data of each channel, extracting color features of the effective image area to serve as target image features, wherein the color features comprise at least one of the number of colors and distribution proportion thereof, the dispersion degree of the colors, color uniformity, brightness, average color and main color, and/or calculating a gray level symbiotic matrix of the effective image area, extracting texture features of the effective image area to serve as target image features, wherein the texture features comprise at least one of contrast, correlation, entropy and uniformity, performing contour analysis on the effective image area to obtain geometric features of the effective image area to serve as target image features, and/or performing contour analysis on the effective image area to obtain the effective image area to serve as spectral features, wherein the spectral features comprise at least one of the shape, size and surface area, and/or obtaining the spectral features of the effective image area to include at least one of the spectral features, namely the spectral features of the target image and the spectral features.
Optionally, before the target image features are input into the trained prediction model and the actual grade of the litchi is output through the prediction model, the method further comprises the steps of identifying and marking the grade of the litchi in advance to serve as a marking grade, inputting the target image features into an initial prediction model, analyzing contribution degree of the target image features to a prediction result according to the marking grade, removing features with low contribution degree or negative influence on the prediction result, and retaining the features with high contribution degree to obtain the trained prediction model.
Optionally, the step of inputting the target image features into a trained prediction model and outputting the actual grade of the litchi through the prediction model comprises the steps of inputting the target image features into the trained prediction model to obtain the pre-judging grade output by each decision tree in the prediction model, and screening the pre-judging grade with the largest quantity from the pre-judging grades to serve as the actual grade of the litchi.
Optionally, after the target image features are input into the trained prediction model and the actual grade of the litchi is output through the prediction model, the method further comprises feeding back the actual grade of the litchi in real time.
The application provides a litchi nondestructive sorting device, which comprises an image acquisition module, an image processing module, a characteristic extraction module and a grade determination module, wherein the image acquisition module is used for acquiring multispectral images of litchi, the multispectral images comprise at least two of visible light images, infrared light images and ultraviolet light images, the image processing module is used for inputting the multispectral images into an image segmentation network to carry out segmentation processing to obtain an effective image area of the litchi, the characteristic extraction module is used for extracting target image characteristics of the effective image area, the grade determination module is used for inputting the target image characteristics into a trained prediction model, and outputting the actual grade of the litchi through the prediction model.
Optionally, the feature extraction module is specifically configured to extract initial image features of the effective image area, and screen target image features from the initial image features, where the number of the initial image features is greater than the number of the target image features.
Optionally, the image processing module is specifically configured to convert the multispectral image into a gray image, perform binarization processing on the gray image to obtain a binarized image, perform morphological operation on the binarized image to obtain a morphological image, where the morphological operation includes at least one of expansion, corrosion, open operation and close operation, and extract edges of the morphological image by an edge detection algorithm to obtain an effective image area of the litchi.
The feature extraction module is specifically configured to separate RGB channels from the effective image area, extract RGB image data of red, green and blue channels respectively, perform statistical calculation on the RGB image data of each channel, extract color features of the effective image area as target image features, where the color features include at least one of a number of colors and a distribution ratio thereof, a dispersion degree of the colors, color uniformity, brightness, an average color and a main color, and/or calculate a gray level co-occurrence matrix of the effective image area, extract texture features of the effective image area as target image features, where the texture features include at least one of a contrast ratio, a correlation, an entropy and a uniformity, and/or perform contour analysis on the effective image area to obtain geometric features of the effective image area as target image features, where the geometric features include at least one of a shape, a size and a surface area, and/or analyze the effective image area to obtain a gray level co-occurrence matrix of the effective image area, where the texture features include at least one of a refractive index, a spectral feature and a spectral feature.
Optionally, the litchi nondestructive sorting device further comprises a model training module, wherein the model training module is used for identifying and marking the grade of the litchi in advance to be used as a marking grade, inputting the target image characteristics into an initial prediction model, analyzing the contribution degree of the target image characteristics to a prediction result according to the marking grade, removing the characteristics with low contribution degree or negative influence on the prediction result, and retaining the characteristics with high contribution degree to obtain a trained prediction model.
Optionally, the grade determining module is specifically configured to input the target image feature into a trained prediction model to obtain a pre-judging grade output by each decision tree in the prediction model, and screen out the pre-judging grade with the largest number as an actual grade of the litchi.
Optionally, the litchi nondestructive sorting device further comprises a feedback module for feeding back the actual grade of the litchi in real time.
The third aspect of the present application also provides a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor, when executing the computer program, implementing the steps of the litchi non-destructive sorting method as described in any one of the above.
The fourth aspect of the present application also provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the litchi non-destructive sorting method as set forth in any one of the preceding claims.
The litchi nondestructive sorting method, the device, the computer equipment and the computer readable storage medium have the following beneficial effects that the litchi quality is sorted in a grade mode based on multispectral imaging and deep learning litchi quality prediction models, and the problems of poor timeliness, high cost, strong subjectivity and the like in the prior art when the litchi quality is detected manually can be solved. By the artificial intelligence-based litchi nondestructive sorting method, litchi can be rapidly, noninvasively and accurately predicted in quality, so that the efficiency and sorting precision of a production line are improved. In addition, by acquiring at least two of the multi-spectral images, such as the visible light image, the infrared light image and the ultraviolet light image, a plurality of groups of spectral images can be obtained, and data can be provided for subsequent predictive analysis in a multi-angle and mutually-cooperated manner, so that the litchi grade result output by the predictive model is more accurate and reliable, the efficiency and accuracy of litchi quality assessment are remarkably improved, and the high-quality development of the litchi intelligent sorting industry is promoted.
Drawings
FIG. 1 is a schematic diagram of a non-destructive sorting method for litchi in an embodiment of the present application;
FIG. 2 is a block diagram of a litchi nondestructive sorting device in an embodiment of the application;
Fig. 3 is a block diagram schematically illustrating a structure of a computer device according to an embodiment of the present application.
The achievement of the objects, functional features and advantages of the present application will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
Referring to fig. 1, the embodiment of the application provides a litchi nondestructive sorting method, which comprises the steps of S101, S102, S103, S104, wherein the step is to obtain a multispectral image of litchi, the step is to input the multispectral image into an image segmentation network for segmentation processing to obtain an effective image area of the litchi, the step is to extract target image characteristics of the effective image area, and the step is to input the target image characteristics into a trained prediction model and output actual grades of the litchi through the prediction model.
According to the litchi nondestructive sorting method provided by the embodiment of the application, the litchi quality is sorted in a grade mode based on multispectral imaging and deep learning litchi quality prediction models, so that the problems of poor timeliness, high cost, strong subjectivity and the like in the prior art when the litchi quality is detected manually can be solved. By the artificial intelligence-based litchi nondestructive sorting method, litchi can be rapidly, noninvasively and accurately predicted in quality, so that the efficiency and sorting precision of a production line are improved. In addition, by acquiring at least two of the multi-spectral images, such as the visible light image, the infrared light image and the ultraviolet light image, a plurality of groups of spectral images can be obtained, and data can be provided for subsequent predictive analysis in a multi-angle and mutually-cooperated manner, so that the litchi grade result output by the predictive model is more accurate and reliable, the efficiency and accuracy of litchi quality assessment are remarkably improved, and the high-quality development of the litchi intelligent sorting industry is promoted.
In step S101, a multispectral image of litchi is acquired.
In some embodiments, the multispectral image includes at least one of a visible light image, an infrared light image, and an ultraviolet light image. In other embodiments, the multispectral image includes at least two of a visible light image, an infrared light image, and an ultraviolet light image. In still other embodiments, the multispectral image includes a visible light image, an infrared light image, and an ultraviolet light image. The visible light image refers to an image obtained by irradiating litchi by a visible light source (such as an RGB light source, namely, red, green and blue light sources) and photographing by a camera with a visible light imaging system, the infrared light image refers to an image obtained by irradiating litchi by an infrared light source (such as a near infrared light source and/or a far infrared light source) and photographing by a camera with an infrared imaging system, and the ultraviolet light image refers to an image obtained by irradiating litchi by an ultraviolet light source (such as a near ultraviolet light source and/or a far ultraviolet light source) and photographing by a camera with an ultraviolet imaging system. In some embodiments, the camera may have an imaging system with only one type of light source, or may have an imaging system with two or more types of light sources (e.g., a multispectral imaging system with visible, infrared, and ultraviolet light simultaneously).
The method comprises the steps of capturing color and texture information of the surface of the litchi through a visible light image, identifying possible defects or maturity by capturing internal structures and characteristics of the litchi through an infrared light image, exciting fluorescent reaction of the surface and the inside of the litchi through ultraviolet light, and revealing possible quality problems of the microstructure and the inside of the surface of the litchi through an ultraviolet light image. Therefore, when the multispectral image of the litchi is obtained and comprises at least two of a visible light image, an infrared light image and an ultraviolet light image, a plurality of groups of the multispectral images can be obtained at the same time, and the surface and the internal characteristics of the litchi can be extracted at the same time, so that data are provided for subsequent predictive analysis in a multi-angle and mutual cooperation manner, the litchi grade result output by the predictive model is more accurate and reliable, the efficiency and the accuracy of litchi quality assessment are remarkably improved, and the high-quality development of the litchi intelligent sorting industry is promoted.
In order to make the multi-spectral image of litchi obtained in step S101 clearer and more accurate, a plurality of light sources may be uniformly arranged around the litchi before step S101, wherein the light sources include at least two of visible light sources, infrared light sources and ultraviolet light sources. In some embodiments, the number of the visible light sources may be more than two, for example, four, and the visible light sources are respectively disposed on the upper, lower, left and right surfaces of the litchi. In other embodiments, the number of the infrared light sources may be more than two, for example, four, and the infrared light sources are respectively disposed on the upper, lower, left and right surfaces of the litchi. In still other embodiments, the number of ultraviolet light sources may be more than two, for example, two, and the ultraviolet light sources may be respectively disposed on the upper surface and the lower surface of the litchi. The positions of the visible light source, the infrared light source and the ultraviolet light source can be completely staggered, partially overlapped and completely overlapped, and particularly can be adjusted according to actual needs so as to ensure that various light sources can uniformly cover the surface of litchi.
Further, it may further include, prior to step S101, capturing at least two of the visible light image, the infrared light image, and the ultraviolet light image simultaneously using a multispectral camera. In some embodiments, a multispectral filter may be provided in the multispectral camera. Through multispectral camera, can shoot simultaneously and obtain more than two kinds of spectral image to through multispectral filter's setting, can ensure that the image of each kind of light source can gather alone, simplify the image acquisition process, and make the image of gathering more clear.
In step S102, the multispectral image is input into an image segmentation network for segmentation processing, so as to obtain an effective image region of litchi. In some embodiments, the image segmentation network may be a deep learning based Fun, unet, unet ++, or the like image segmentation network. In some embodiments, the multispectral image is input into an image segmentation network for segmentation processing to remove invalid areas such as background or other sundries, and the valid areas belonging to the litchi part are identified and reserved, so that the valid image areas of the litchi are obtained. Therefore, through the image segmentation network, images with different spectrums can be fused and analyzed, and the accuracy of removing the background or other impurities can be improved, so that the accuracy and the recognition efficiency of the effective image area recognition of the litchi are improved.
In some embodiments, the multispectral image can be a color image, the multispectral image is input into an image segmentation network for segmentation processing to obtain an effective image area of the litchi, and the multispectral image can be converted into a gray-scale image, the gray-scale image is subjected to binarization processing to obtain a binarized image, and edges of the binarized image are extracted through an edge detection algorithm to obtain the effective image area of the litchi. Therefore, the obtained binarized image can be better used for extracting the effective image area of the litchi by converting the multispectral image into a gray level image and performing binarization processing.
In some embodiments, the multispectral image can be a color image, the multispectral image is input into an image segmentation network for segmentation processing to obtain an effective image area of the litchi, and the multispectral image can be converted into a gray image, the gray image is subjected to binarization processing to obtain a binarized image, morphological operation is performed on the binarized image to obtain a morphological image, and edges of the morphological image are extracted through an edge detection algorithm to obtain the effective image area of the litchi. Therefore, the multispectral image is converted into the gray level image and binarized, morphological operation can be carried out on the image more effectively, and the edge and related detail characteristics of the litchi can be extracted better through the morphological operation, so that a more accurate effective image area is obtained.
In some embodiments, the morphological operations may include at least one of dilation, erosion, open operation, and closed operation. The method comprises the steps of expanding a white area in an image, filling small holes, connecting adjacent areas, reducing the white area in the image, removing small noise points, separating the adjacent areas, performing operation of firstly etching and then expanding, wherein the operation of opening is mainly used for removing the small noise in the image, and the operation of closing is firstly expanding and then etching, and is mainly used for filling the small holes in the image and connecting the adjacent areas. Therefore, through the morphological operation, the image before processing can be further perfected (such as filling small holes or removing noise), so that the processed image is clearer and more accurate.
In some embodiments, the multispectral image can be a gray level image directly, and the multispectral image is input into an image segmentation network for segmentation processing to obtain an effective image area of the litchi, and the method can comprise the steps of performing binarization processing on the gray level image to obtain a binarized image. In other embodiments, the method may further comprise performing morphological operations on the binarized image to obtain a morphological image. In still other embodiments, the method can further comprise extracting edges of the binarized image or the morphological image by an edge detection algorithm to obtain an effective image area of the litchi. Therefore, as the multispectral image is originally a gray level image, the gray level processing step is not needed, and the multispectral image can be directly subjected to binarization processing, so that the effective image area of the litchi is obtained by extracting the edges of the binarized image. The method can further perform morphological operation on the binarized image, and the edges and related detail features of the litchi can be better extracted through morphological operation, so that a more accurate effective image area is obtained.
In some embodiments, the edge detection algorithm may include a multi-level edge detection algorithm, such as a Canny algorithm, a Sobel algorithm, or a Laplace algorithm, or the like. The accuracy of extracting the edges and related detail features of the litchi can be further improved through a multi-stage edge detection algorithm, so that the obtained effective image area is more accurate and reliable.
In step S103, target image features of the effective image area are extracted.
In some embodiments, extracting the target image features of the effective image region may include extracting initial image features of the effective image region and screening the target image features from the initial image features. Wherein the number of initial image features may be the same as the number of target image features. That is, in some cases, if the extracted initial image feature data are all better (for example, all satisfy the preset condition interval), the initial image feature may be directly used as the target image feature, or all the initial image features may be selected as the target image feature during screening.
In other embodiments, the number of initial image features may be greater than the number of target image features. In this case, data with poor results (e.g., data that does not satisfy the preset condition interval) in the initial image feature may be excluded, and data with good results (e.g., data that satisfies the preset condition interval) may be screened out therefrom as the target image feature. Illustratively, litchi X of grade A has a satisfactory target image characteristic of abc (where d in the initial image characteristic does not meet the preset condition), while litchi Y of grade A has a satisfactory target image characteristic of acd (where b in the initial image characteristic does not meet the preset condition), but both grades are A and should be classified as the same quality grade, if the target image characteristic is fixed to abc, it will necessarily result in litchi Y being ultimately determined as not belonging to the quality grade of litchi X, and if the target image characteristic is fixed to acd, it will necessarily result in litchi X being ultimately determined as not belonging to the quality grade of litchi Y. In this case, the result of the final determination is not in agreement with the actual situation (litchi X and Y belong to the same quality class).
Therefore, by adding the step of screening the target image features from the initial image features, more excellent target image features can be adaptively screened in the presence of a plurality of optional features (features affecting the litchi quality), so that the consistency of the grading of litchis with different features but identical quality can be ensured.
In some embodiments, extracting the target image features of the effective image region may include performing RGB channel separation on the effective image region, extracting RGB image data of red, green, and blue channels, respectively, and performing statistical calculation on the RGB image data of each channel to extract color features of the effective image region as the target image features. In other embodiments, the color characteristics may include at least one of the number of colors and their distribution ratio, the degree of dispersion of the colors, color uniformity, brightness, average color, and dominant color. The color data is extracted through RGB channels, and statistical parameters such as a histogram, a mean value, a standard deviation and the like of each channel are calculated to serve as color features (namely target image features), and color and texture information of the litchi surface can be obtained through the color features so as to reflect the appearance quality of the litchi.
In some embodiments, extracting the target image features of the effective image region may include calculating a Gray-Level Co-occurrence Matrix, GLCM of the effective image region and extracting texture features of the effective image region as the target image features. In other embodiments, the texture features may include at least one of contrast, correlation, entropy, and uniformity. The texture features of the litchi can be extracted by calculating and analyzing the gray level co-occurrence matrix, so as to be used for evaluating the surface structure and the texture of the litchi.
In some embodiments, extracting the target image features of the effective image region may include performing contour analysis on the effective image region to obtain geometric features of the effective image region as the target image features. In other embodiments, the geometric features may include at least one of shape, size, and surface area. The surface area of the litchi can be calculated by extracting the outline of the effective image area. Through contour analysis, the geometric characteristics of litchi can be obtained, so that the shape, the size, the weight and the like of the litchi can be evaluated conveniently.
In some embodiments, extracting the target image features of the effective image region may include performing spectral analysis on the effective image region to obtain spectral features of the effective image region as the target image features. In other embodiments, the spectral features include at least one of reflectivity, refractive index, and absorptivity. The surface topography of litchi can be further reflected by the spectral features.
In some embodiments, extracting the target image features of the active image region may include both of the above-described approaches, such that the target image features may include at least two of color features, texture features, geometric features, and spectral features. Therefore, at least two of the color features, texture features, geometric features and spectral features of the litchi can be fused together, and the accuracy of the litchi sorting result can be further improved through a multilayer feature fusion technology.
In step S104, the target image features are input into the trained prediction model, and the actual level of litchi is output through the prediction model. The predictive model may include, among other things, a deep learning-based predictive model, such as a random forest model, XGBoost (distributed gradient-enhanced library) model, or LightGBM (lightweight gradient hoist) model, among others. Through the prediction model based on deep learning, large-scale data and characteristics can be processed, and the analysis efficiency and the judgment accuracy are higher, so that the efficiency of the litchi sorting process and the accuracy and reliability of results are improved.
In some embodiments, the predictive model is N, where N is an odd number greater than or equal to 3. In some embodiments, inputting the target image features into the trained prediction models and outputting the actual grade of the litchi through the prediction models may include inputting the target image features into the N trained prediction models respectively, outputting the predicted grade of the litchi by each prediction model respectively, and screening out the most number of predicted grades from the predicted grades to serve as the actual grade of the litchi. Therefore, the accuracy of the litchi grade sorting result can be further improved through the collaborative prediction of the plurality of prediction models, so that the reliability of the sorting method is improved.
In some embodiments, inputting the target image features into the trained prediction model and outputting the actual grade of the litchi through the prediction model may include inputting the target image features into the trained prediction model to obtain the pre-judgment grade output by each decision tree in the prediction model, and screening out the most number of pre-judgment grades from the pre-judgment grades to serve as the actual grade of the litchi. Therefore, through the arrangement of a plurality of decision trees, the pre-judging grade of the litchi can be obtained in multiple angles, and the actual grade of the litchi is determined through comparison and analysis, so that the accuracy and the reliability of litchi grade sorting are improved.
In some embodiments, before inputting the target image features into the trained prediction model and outputting the actual grade of the litchi through the prediction model, the method further comprises the steps of identifying and marking the grade of the litchi in advance to be used as a marking grade, inputting the target image features into the initial prediction model, analyzing the contribution degree of the target image features to the prediction result according to the marking grade, removing the features with low contribution degree or negative effects on the prediction result, and retaining the features with high contribution degree to obtain the trained prediction model. Therefore, a large amount of data can be input into the prediction model in advance, so that the prediction model can determine the correlation between each parameter and the litchi quality through statistical analysis and a machine learning algorithm, and therefore the characteristic parameters which have obvious influence on the litchi quality are screened out, and the accurate and reliable grading result is ensured.
It should be noted that the contribution degree of the target image features to the prediction result may be set according to the actual requirement. For example, when the influence of a certain target image feature on the prediction result reaches a preset standard value or more, the contribution degree of the target image feature on the prediction result is considered to be high, when the influence of the certain target image feature on the prediction result is considered to be below the preset standard value, the contribution degree of the target image feature on the prediction result is considered to be low, and when the influence of the certain target image feature on the prediction result is considered to be negative (i.e., the result obtained according to the data of the target image feature shows that the litchi grade is A, but the actual level (e.g., the artificially marked level) of the actual litchi is B), the target image feature is considered to have a negative influence on the prediction result. The preset standard value can be adjusted according to actual requirements, for example, 10%, 20%, 30%, 50% and the like.
In some embodiments, pre-identifying and marking the level of litchi as a marking level may include manually identifying and marking the level of litchi as a marking level. Therefore, the method can be manually identified in advance and used as data for training the prediction model, so that the sorting accuracy of the prediction model is improved.
In other embodiments, analyzing the contribution of the target image feature to the prediction based on the marking level may include analyzing the contribution of the target image feature to the prediction based on the marking level using SHAP values (SHAPLEY ADDITIVE exPlanations, saprolimus plus and interpretation). The SHAP value is used for analyzing the contribution degree of each target image feature to the prediction result, and specifically, the method can be used for analyzing the global feature importance, namely, displaying the global feature importance, showing which features have the greatest influence on model prediction as a whole, (2) single prediction interpretation, namely, single sample prediction is explained, and the contribution of each feature to the prediction is shown, and (3) feature interaction analysis, namely, interaction among the features is analyzed, and the performance of a certain feature under different conditions is known.
In some embodiments, based on analysis results, such as SHAP value analysis results, the predictive model may be optimized by (1) feature selection-removing features with low contribution or negative impact on prediction, preserving and accentuated analysis of features with high contribution, and (2) model tuning-tuning model parameters based on distribution and interaction of important features, optimizing model performance. Therefore, a better trained prediction model is obtained, so that the litchi sorting grade output by the prediction model is more accurate and reliable.
In some embodiments, after inputting the target image features into the trained prediction model and outputting the actual grade of the litchi through the prediction model, real-time feedback of the actual grade of the litchi can be further included. For example, the actual grade of the litchi can be fed back to the transmission device in real time, and the transmission device can transmit the litchi to the corresponding channel according to the actual grade of the litchi, so as to realize the litchi sorting effect.
In some embodiments, after inputting the target image features into the trained predictive model, outputting the actual grade of the litchi by the predictive model may further include detecting whether the actual grade of the litchi is accurate. In some embodiments, detecting whether the actual level of litchi is accurate may include detecting whether the actual level of litchi is accurate when a preset time or a preset count is reached. For example, after litchi is sorted for a certain time (for example, 1 h) or a certain number (for example, 100) by adopting the sorting method, one or more sorted litchis are manually identified, whether the result of the manual identification accords with the predicted actual grade of the litchi is judged, if so, the litchi is accurate, and if not, the litchi is inaccurate.
In some embodiments, sorting can be continued when the actual grade of the litchi is detected to be accurate, and when the actual grade of the litchi is detected to be inaccurate, after detecting whether the actual grade of the litchi is accurate, the method can further comprise inputting a result of manual identification into a prediction model, and further training and optimizing the prediction model. Therefore, the prediction model is continuously self-optimized through deep learning, and the sorting standard and parameters are continuously adjusted so as to improve the intelligent level of sorting, thereby improving the accuracy and the sorting efficiency of the litchi sorting result.
Referring to fig. 2, the embodiment of the application further provides a litchi nondestructive sorting device, which comprises an image acquisition module 201, an image processing module 202, a feature extraction module 203 and a grade determination module 204, wherein the image acquisition module is used for acquiring multispectral images of litchi, the multispectral images comprise at least two of visible light images, infrared light images and ultraviolet light images, the image processing module 202 is used for inputting the multispectral images into an image segmentation network to carry out segmentation processing to obtain an effective image area of the litchi, the feature extraction module 203 is used for extracting target image features of the effective image area, the grade determination module 204 is used for inputting the target image features into a trained prediction model, and outputting the actual grade of the litchi through the prediction model.
The litchi nondestructive sorting device provided by the embodiment of the application realizes the grade sorting of litchi quality by a litchi quality prediction model based on multispectral imaging and deep learning, and can solve the problems of poor timeliness, high cost, strong subjectivity and the like in the prior art when the litchi quality is detected manually. By the artificial intelligence-based litchi nondestructive sorting method, litchi can be rapidly, noninvasively and accurately predicted in quality, so that the efficiency and sorting precision of a production line are improved. In addition, by acquiring at least two of the multi-spectral images, such as the visible light image, the infrared light image and the ultraviolet light image, a plurality of groups of spectral images can be obtained, and data can be provided for subsequent predictive analysis in a multi-angle and mutually-cooperated manner, so that the litchi grade result output by the predictive model is more accurate and reliable, the efficiency and accuracy of litchi quality assessment are remarkably improved, and the high-quality development of the litchi intelligent sorting industry is promoted.
In some embodiments, the feature extraction module 203 may be specifically configured to extract initial image features of the effective image area, and screen target image features from the initial image features. Wherein the number of initial image features may be the same as the number of target image features. That is, in some cases, if the extracted initial image feature data are all better (for example, all satisfy the preset condition interval), the initial image feature may be directly used as the target image feature, or all the initial image features may be selected as the target image feature during screening.
In other embodiments, the number of initial image features is greater than the number of target image features. In this case, data with poor results (e.g., data that does not satisfy the preset condition interval) in the initial image feature may be excluded, and data with good results (e.g., data that satisfies the preset condition interval) may be screened out therefrom as the target image feature. By adding the step of screening the target image features from the initial image features, more optimal target image features can be adaptively screened under the condition that a plurality of optional features (features affecting the litchi quality) exist, so that the consistency of the classification of litchis with different features and the same quality can be ensured.
In some embodiments, the image processing module 202 may be specifically configured to convert a multispectral image into a gray-scale image, perform binarization processing on the gray-scale image to obtain a binarized image, perform morphological operations on the binarized image to obtain a morphological image, where the morphological operations include at least one of dilation, erosion, open operation and close operation, and extract edges of the morphological image by an edge detection algorithm to obtain an effective image area of litchi. Therefore, the multispectral image is converted into the gray level image and binarized, morphological operation can be carried out on the image more effectively, and the edge and related detail characteristics of the litchi can be extracted better through the morphological operation, so that a more accurate effective image area is obtained.
In some embodiments, the feature extraction module 203 may be further configured to perform RGB channel separation on the effective image area, extract RGB image data of red, green and blue channels respectively, perform statistical calculation on the RGB image data of each channel, extract color features of the effective image area as target image features, where the color features include at least one of a number of colors and a distribution ratio thereof, a dispersion degree of the colors, color uniformity, brightness, an average color and a main color, and/or calculate a gray co-occurrence matrix of the effective image area, extract texture features of the effective image area as target image features, where the texture features include at least one of contrast, correlation, entropy and uniformity, and/or perform contour analysis on the effective image area to obtain geometric features of the effective image area including at least one of a shape, a size and a surface area, and/or perform spectral analysis on the effective image area to obtain spectral features of the effective image area as target image features, where the spectral features include at least one of a reflectance, a refraction and an absorption.
In some embodiments, extracting the target image features of the active image region may include both of the above-described approaches, such that the target image features may include at least two of color features, texture features, geometric features, and spectral features. Therefore, at least two of the color features, texture features, geometric features and spectral features of the litchi can be fused together, and the accuracy of the litchi sorting result can be further improved through a multilayer feature fusion technology.
In some embodiments, the litchi lossless sorting device further comprises a model training module, wherein the model training module is used for identifying and marking the grade of litchi in advance to be used as a marking grade, inputting the target image characteristics into an initial prediction model, analyzing the contribution degree of the target image characteristics to the prediction result according to the marking grade, removing the characteristics with low contribution degree or negative influence on the prediction result, and retaining the characteristics with high contribution degree to obtain a trained prediction model. Therefore, a large amount of data can be input into the prediction model in advance, so that the prediction model can determine the correlation between each parameter and the litchi quality through statistical analysis and a machine learning algorithm, and therefore the characteristic parameters which have obvious influence on the litchi quality are screened out, and the accurate and reliable grading result is ensured.
In some embodiments, the level determining module 204 may be specifically configured to input the target image features into the trained prediction model to obtain the pre-determined levels output by the decision trees in the prediction model, and select the most number of pre-determined levels from the pre-determined levels to be the actual level of the litchi. Therefore, through the arrangement of a plurality of decision trees, the pre-judging grade of the litchi can be obtained in multiple angles, and the actual grade of the litchi is determined through comparison and analysis, so that the accuracy and the reliability of litchi grade sorting are improved.
In some embodiments, the litchi nondestructive sorting device can further comprise a feedback module for feeding back the actual grade of the litchi in real time. For example, the actual grade of the litchi can be fed back to the transmission device in real time, and the transmission device can transmit the litchi to the corresponding channel according to the actual grade of the litchi, so as to realize the litchi sorting effect.
For specific implementation of each module in the above embodiment of the apparatus, please refer to the above embodiment of the method, and detailed description is omitted herein.
Referring to fig. 3, in an embodiment of the present application, there is further provided a computer device, which may be a server, and an internal structure thereof may be as shown in fig. 3. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected by a system bus. Wherein the computer is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used for communicating with an external terminal through a network connection. Which computer program, when being executed by a processor, carries out the above-mentioned method.
It will be appreciated by those skilled in the art that the architecture shown in fig. 3 is merely a block diagram of a portion of the architecture in connection with the present inventive arrangements and is not intended to limit the computer devices to which the present inventive arrangements are applicable.
The embodiment of the application also provides a computer readable storage medium, on which a computer program is stored, which realizes the litchi nondestructive sorting method when being executed by a processor. It is understood that the computer readable storage medium in this embodiment may be a volatile readable storage medium or a nonvolatile readable storage medium.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium provided by the present application and used in embodiments may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (SYNCHLINK) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the phrase "comprising one does not exclude the presence of other like elements in a process, apparatus, article, or method that comprises the element.
The foregoing description is only of the preferred embodiments of the present application and is not intended to limit the scope of the application, and all equivalent structures or equivalent processes using the descriptions and drawings of the present application or direct or indirect application in other related technical fields are included in the scope of the present application.

Claims (10)

1.一种荔枝无损分选方法，其特征在于，包括：1. A method for non-destructive sorting of litchi, comprising: 获取荔枝的多光谱图像，其中，所述多光谱图像包括可见光图像、红外光图像和紫外光图像中的至少两种；Acquire a multispectral image of litchi, wherein the multispectral image includes at least two of a visible light image, an infrared light image, and an ultraviolet light image; 将所述多光谱图像输入至图像分割网络中进行分割处理，得到所述荔枝的有效图像区域；Inputting the multispectral image into an image segmentation network for segmentation processing to obtain a valid image area of the litchi; 提取所述有效图像区域的目标图像特征；Extracting target image features of the effective image area; 将所述目标图像特征输入至训练后的预测模型中，通过所述预测模型输出所述荔枝的实际等级。The target image features are input into the trained prediction model, and the actual grade of the litchi is output through the prediction model. 2.根据权利要求1所述的荔枝无损分选方法，其特征在于，所述提取所述有效图像区域的目标图像特征，包括：2. The method for lossless sorting of litchi according to claim 1, characterized in that the step of extracting target image features of the effective image area comprises: 提取所述有效图像区域的初始图像特征；Extracting initial image features of the effective image area; 从所述初始图像特征中筛选出目标图像特征，其中，所述初始图像特征的数量大于所述目标图像特征的数量。Target image features are screened out from the initial image features, wherein the number of the initial image features is greater than the number of the target image features. 3.根据权利要求1所述的荔枝无损分选方法，其特征在于，所述将所述多光谱图像输入至图像分割网络中进行分割处理，得到所述荔枝的有效图像区域，包括：3. The method for lossless sorting of litchi according to claim 1, characterized in that the step of inputting the multispectral image into an image segmentation network for segmentation processing to obtain the effective image area of the litchi comprises: 将所述多光谱图像转换为灰度图像；Converting the multispectral image into a grayscale image; 对所述灰度图像进行二值化处理，得到二值化图像；Binarizing the grayscale image to obtain a binary image; 对所述二值化图像进行形态学操作，得到形态学图像，其中，所述形态学操作包括膨胀、腐蚀、开运算和闭运算中的至少一种；Performing a morphological operation on the binary image to obtain a morphological image, wherein the morphological operation includes at least one of dilation, erosion, opening operation and closing operation; 通过边缘检测算法对所述形态学图像的边缘进行提取，得到所述荔枝的有效图像区域。The edge of the morphological image is extracted by an edge detection algorithm to obtain the effective image area of the litchi. 4.根据权利要求1所述的荔枝无损分选方法，其特征在于，所述提取所述有效图像区域的目标图像特征，包括：4. The method for lossless sorting of litchi according to claim 1, characterized in that the step of extracting target image features of the effective image area comprises: 对所述有效图像区域进行RGB通道分离，分别提取红色、绿色和蓝色通道的RGB图像数据，并对各通道的所述RGB图像数据进行统计计算，提取所述有效图像区域的颜色特征，以作为目标图像特征，所述颜色特征包括颜色的数量及其分布比例、颜色的分散程度、颜色均匀性、亮度、平均颜色和主要颜色中的至少一种；和/或，Separating the effective image area into RGB channels, extracting RGB image data of red, green and blue channels respectively, and performing statistical calculation on the RGB image data of each channel to extract color features of the effective image area as target image features, wherein the color features include at least one of the number of colors and their distribution ratio, color dispersion, color uniformity, brightness, average color and main color; and/or, 计算所述有效图像区域的灰度共生矩阵，并提取所述有效图像区域的纹理特征，以作为目标图像特征，所述纹理特征包括对比度、相关性、熵和均匀性中的至少一种；和/或，Calculating the gray level co-occurrence matrix of the effective image area, and extracting the texture features of the effective image area as target image features, wherein the texture features include at least one of contrast, correlation, entropy and uniformity; and/or, 对所述有效图像区域进行轮廓分析，得到所述有效图像区域的几何特征，以作为目标图像特征，所述几何特征包括形状、大小和表面积中的至少一种；和/或，Performing contour analysis on the effective image area to obtain geometric features of the effective image area as target image features, wherein the geometric features include at least one of shape, size and surface area; and/or, 对所述有效图像区域进行光谱分析，得到所述有效图像区域的光谱特征，以作为目标图像特征，所述光谱特征包括反射率、折射率和吸收率中的至少一种。Performing spectral analysis on the effective image area to obtain spectral features of the effective image area as target image features, wherein the spectral features include at least one of reflectivity, refractive index and absorptivity. 5.根据权利要求1所述的荔枝无损分选方法，其特征在于，在所述将所述目标图像特征输入至训练后的预测模型中，通过所述预测模型输出所述荔枝的实际等级之前，还包括：5. The method for non-destructive sorting of litchi according to claim 1, characterized in that before inputting the target image features into the trained prediction model and outputting the actual grade of the litchi through the prediction model, it also comprises: 预先识别并标记所述荔枝的等级，以作为标记等级；Pre-identifying and marking the grade of the litchi as a marked grade; 将所述目标图像特征输入至初始的预测模型中；Inputting the target image features into an initial prediction model; 根据所述标记等级，分析所述目标图像特征对预测结果的贡献度；Analyzing the contribution of the target image features to the prediction result according to the marking level; 去除贡献度低或对预测结果有负面影响的特征，保留贡献度高的特征，以得到训练后的预测模型。Remove features with low contribution or negative impact on the prediction results, and retain features with high contribution to obtain the trained prediction model. 6.根据权利要求1所述的荔枝无损分选方法，其特征在于，所述将所述目标图像特征输入至训练后的预测模型中，通过所述预测模型输出所述荔枝的实际等级，包括：6. The method for non-destructive sorting of litchi according to claim 1, characterized in that the step of inputting the target image features into a trained prediction model and outputting the actual grade of the litchi through the prediction model comprises: 将所述目标图像特征输入至训练后的预测模型中，得到所述预测模型中各个决策树输出的预判等级；Inputting the target image features into the trained prediction model to obtain the prediction level output by each decision tree in the prediction model; 从所述预判等级中，筛选出数量最多的预判等级，以作为所述荔枝的实际等级。From the predicted grades, the predicted grades with the largest number are screened out to serve as the actual grades of the litchi. 7.根据权利要求1所述的荔枝无损分选方法，其特征在于，在所述将所述目标图像特征输入至训练后的预测模型中，通过所述预测模型输出所述荔枝的实际等级之后，还包括：7. The method for non-destructive sorting of litchi according to claim 1, characterized in that after inputting the target image features into the trained prediction model and outputting the actual grade of the litchi through the prediction model, it further comprises: 实时反馈所述荔枝的所述实际等级。The actual grade of the litchi is fed back in real time. 8.一种荔枝无损分选装置，其特征在于，包括：8. A non-destructive litchi sorting device, comprising: 图像获取模块，用于获取荔枝的多光谱图像，其中，所述多光谱图像包括可见光图像、红外光图像和紫外光图像中的至少两种；An image acquisition module, used to acquire a multispectral image of litchi, wherein the multispectral image includes at least two of a visible light image, an infrared light image, and an ultraviolet light image; 图像处理模块，用于将所述多光谱图像输入至图像分割网络中进行分割处理，得到所述荔枝的有效图像区域；An image processing module, used for inputting the multispectral image into an image segmentation network for segmentation processing to obtain a valid image area of the litchi; 特征提取模块，用于提取所述有效图像区域的目标图像特征；A feature extraction module, used to extract target image features in the effective image area; 等级确定模块，用于将所述目标图像特征输入至训练后的预测模型中，通过所述预测模型输出所述荔枝的实际等级。The grade determination module is used to input the target image features into the trained prediction model and output the actual grade of the litchi through the prediction model. 9.一种计算机设备，包括存储器和处理器，所述存储器中存储有计算机程序，其特征在于，所述处理器执行所述计算机程序时实现权利要求1至7中任一项所述荔枝无损分选方法的步骤。9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein the processor implements the steps of the lychee non-destructive sorting method according to any one of claims 1 to 7 when executing the computer program. 10.一种计算机可读存储介质，其上存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现权利要求1至7中任一项所述的荔枝无损分选方法的步骤。10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the lychee lossless sorting method according to any one of claims 1 to 7 are implemented.
CN202411133719.2A 2024-08-19 2024-08-19 Litchi nondestructive sorting method and device, storage medium and computer equipment Pending CN119049041A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202411133719.2A CN119049041A (en) 2024-08-19 2024-08-19 Litchi nondestructive sorting method and device, storage medium and computer equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202411133719.2A CN119049041A (en) 2024-08-19 2024-08-19 Litchi nondestructive sorting method and device, storage medium and computer equipment

Publications (1)

Publication Number Publication Date
CN119049041A true CN119049041A (en) 2024-11-29

Family

ID=93577246

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202411133719.2A Pending CN119049041A (en) 2024-08-19 2024-08-19 Litchi nondestructive sorting method and device, storage medium and computer equipment

Country Status (1)

Country Link
CN (1) CN119049041A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119619026A (en) * 2025-02-17 2025-03-14 陕西华乐宝生物科技有限公司 A nondestructive detection imaging control system for fruit and vegetable quality
CN119672006A (en) * 2024-12-10 2025-03-21 永德智辛觅科技有限公司 A tea quality detection method, system, device and medium based on big data
CN119845962A (en) * 2025-02-18 2025-04-18 陕西果业科技集团有限公司 Real-time on-line detection method and system for appearance defects of fruits

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101059452A (en) * 2007-05-29 2007-10-24 浙江大学 Fruit quality damage-free detection method and system based on multiple spectral imaging technique
CN113989639A (en) * 2021-10-20 2022-01-28 华南农业大学 Litchi disease automatic identification method and device based on hyperspectral image analysis and processing method
CN114112932A (en) * 2021-11-08 2022-03-01 南京林业大学 Hyperspectral detection method and sorting equipment for maturity of oil-tea camellia fruits based on deep learning
CN115713762A (en) * 2022-11-23 2023-02-24 河南讯飞人工智能科技有限公司 Fruit grading method, fruit grading device, electronic equipment, sorting system and storage medium
CN118446969A (en) * 2024-04-25 2024-08-06 华南农业大学 Litchi phenotypic parameter detection method and system based on image processing and three-dimensional point cloud

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101059452A (en) * 2007-05-29 2007-10-24 浙江大学 Fruit quality damage-free detection method and system based on multiple spectral imaging technique
CN113989639A (en) * 2021-10-20 2022-01-28 华南农业大学 Litchi disease automatic identification method and device based on hyperspectral image analysis and processing method
CN114112932A (en) * 2021-11-08 2022-03-01 南京林业大学 Hyperspectral detection method and sorting equipment for maturity of oil-tea camellia fruits based on deep learning
CN115713762A (en) * 2022-11-23 2023-02-24 河南讯飞人工智能科技有限公司 Fruit grading method, fruit grading device, electronic equipment, sorting system and storage medium
CN118446969A (en) * 2024-04-25 2024-08-06 华南农业大学 Litchi phenotypic parameter detection method and system based on image processing and three-dimensional point cloud

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
彭彦昆 等: "农产品智能检测与分级", 30 April 2024, 华中科技大学出版社, pages: 101 - 104 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119672006A (en) * 2024-12-10 2025-03-21 永德智辛觅科技有限公司 A tea quality detection method, system, device and medium based on big data
CN119672006B (en) * 2024-12-10 2025-06-24 广西横县张一元饮品有限公司 A tea quality detection method, system, device and medium based on big data
CN119619026A (en) * 2025-02-17 2025-03-14 陕西华乐宝生物科技有限公司 A nondestructive detection imaging control system for fruit and vegetable quality
CN119845962A (en) * 2025-02-18 2025-04-18 陕西果业科技集团有限公司 Real-time on-line detection method and system for appearance defects of fruits

Similar Documents

Publication Publication Date Title
CN111862064B (en) Silver wire surface flaw identification method based on deep learning
CN111667455B (en) AI detection method for brushing multiple defects
CN117152161B (en) Shaving board quality detection method and system based on image recognition
US9740957B2 (en) Learning pixel visual context from object characteristics to generate rich semantic images
CN119049041A (en) Litchi nondestructive sorting method and device, storage medium and computer equipment
CN115457037B (en) Raw material quality inspection method for donkey-hide gelatin product production
CN108181316B (en) Bamboo strip defect detection method based on machine vision
CN107607554A (en) A kind of Defect Detection and sorting technique of the zinc-plated stamping parts based on full convolutional neural networks
CN111915572A (en) A deep learning-based self-adaptive gear pitting quantitative detection system and method
CN113177924A (en) Industrial production line product flaw detection method
CN118014933B (en) A defect detection and identification method and device based on image detection
Wah et al. Analysis on feature extraction and classification of rice kernels for Myanmar rice using image processing techniques
CN111667475A (en) Machine vision-based Chinese date grading detection method
CN113610831A (en) Wood defect detection method based on computer image technology and transfer learning
CN113295690A (en) Rapid discrimination method of strawberry ripeness based on machine learning
Sun et al. A novel method for multi-feature grading of mango using machine vision
CN119338763A (en) Fresh flower freshness evaluation system based on visual detection
CN119274001B (en) Quantitative recognition method of test paper color change based on improved MSRCR and random forest
CN119599994B (en) An intelligent visual inspection system and method for non-standard automated production lines
CN117853783B (en) A single board defect recognition method and system based on deep learning
CN118864845A (en) A method for identifying and segmenting characteristic regions of optical microscope images
CN117893831A (en) A banana maturity classification method based on semantic segmentation
CN114937042B (en) Plastic product quality evaluation method based on machine vision
CN117392461A (en) Silkworm cocoon picture classification method and system based on ResNet18 and attention mechanism
Bakar et al. GLCM-Based Feature Extraction and CNN Approach for Fruit Freshness Detection

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination