CN119715468A

CN119715468A - Spectral index construction method and device for detecting tea leaves

Info

Publication number: CN119715468A
Application number: CN202411741467.1A
Authority: CN
Inventors: 段丹丹; 赵春江
Original assignee: Research Center of Information Technology of Beijing Academy of Agriculture and Forestry Sciences
Current assignee: Research Center of Information Technology of Beijing Academy of Agriculture and Forestry Sciences
Priority date: 2024-11-29
Filing date: 2024-11-29
Publication date: 2025-03-28

Abstract

The present invention provides a method and device for constructing a spectral index for detecting tea leaves, comprising: obtaining multispectral reflectance data at the scale of tea leaves canopy; constructing a multispectral full-channel difference vegetation index based on the multispectral using the multispectral reflectance data at the scale of tea leaves canopy; and constructing a spectral index for detecting tea leaves based on the full-channel difference vegetation index. By constructing a multispectral full-channel difference vegetation index based on the multispectral using the acquired multispectral reflectance data, changes in physiological parameters at different vertical structural positions of the tea leaves canopy can be more sensitively reflected. This index improves the meticulousness and accuracy of detection, and helps to discover subtle differences in tea leaves growth. By constructing a spectral index for detecting tea leaves based on the full-channel difference vegetation index, the spectral index can more accurately reflect the growth status and nutritional status of tea leaves, and improve the accuracy of detection.

Description

Spectral index construction method and device for detecting tea leaves

Technical Field

The invention relates to the technical field of remote sensing, in particular to a spectral index construction method and device for detecting tea.

Background

Tea is a popular beverage in the world, and its quality and yield are directly related to the nutritional status of fresh tea leaves. Nitrogen is an essential nutrient element for plant growth and is a key index for measuring the nutrition state of plants. During the growth of tea trees, the nitrogen content has a critical effect on the quality and yield of tea. Meanwhile, chlorophyll is taken as green plant to carry out photosynthesis, and the content of chlorophyll plays a role in fresh tea leaves, and is directly related to the synthesis of tea quality components and other organic matters. Therefore, the chlorophyll content and nitrogen content information of the tea are accurately mastered, and the method has extremely important significance for making scientific and reasonable tea garden cultivation management measures and improving tea quality and yield.

However, the traditional chlorophyll content and nitrogen content measuring method, such as laboratory physicochemical detection, is complex in operation, time-consuming and labor-consuming, and high in cost, and is difficult to realize large-scale popularization and application. In addition, the methods are usually destructive, can cause irreversible damage to the tea samples, have hysteresis in detection results, and are difficult to reflect the actual growth conditions of the tea in time.

In recent years, as a novel nondestructive testing technology, a spectrum measurement technology has received a great deal of attention due to the rapid and accurate characteristics. Among them, hyperspectral measurement techniques are applied in various fields with their high-precision characteristics. However, hyperspectral devices are expensive to sell, limiting the possibilities for their widespread use. Therefore, the multispectral technology has been developed as an improved version of the hyperspectral technology, the equipment cost is greatly reduced, the accuracy of data is ensured to the greatest extent, and the multispectral technology is gradually the preferred scheme in various fields.

Although multispectral techniques have many advantages, in previous studies, such new techniques for non-destructive measurement are based on point-to-point measurement, such as using SPAD and other non-imaging spectrometers, and the measurement efficiency is relatively low. In addition, when performing spectroscopic measurements on the surface scale, the detection of physiological parameters is often not very specific for targets in different vertical structural positions (e.g. tender and strong leaves of tea). Because the tender leaves and the strong leaves have differences in chlorophyll content and the like, the physiological parameters of the tender leaves and the strong leaves are detected separately, so that more accurate results are obtained.

In view of the limitations of the traditional detection method and the advantages and disadvantages of the spectrum measurement technology, it is particularly important to find a rapid, accurate, economical and practical detection method for the nitrogen content and the chlorophyll content of tea leaves. With the rapid development of remote sensing technology, the application of spectrum nondestructive detection technology to nondestructive and rapid detection of physiological and biochemical components of tea has become an important content for current tea growth monitoring and nutrition diagnosis. The application of the technology provides more accurate and timely information support for the establishment of cultivation management measures of tea gardens, and is hopeful to promote the further improvement of tea quality and yield.

Disclosure of Invention

The invention provides a method and a device for constructing a spectrum index for detecting tea, which are used for solving the defect that targets at different vertical structure positions are not detected finely enough in the prior art and improving the accuracy of multi-spectrum detection of the tea. The technical scheme provided by the invention is as follows:

in a first aspect, the present invention provides a method of spectral index construction for detecting tea leaves, comprising:

Obtaining multispectral reflectivity data of tea canopy dimensions;

Constructing a multispectral-based full-channel difference vegetation index by utilizing multispectral reflectivity data of the tea canopy scale;

And constructing a spectrum index for detecting tea based on the full-channel difference vegetation index.

Optionally, the acquiring the spectral reflectance data of the tea canopy scale includes:

Acquiring multispectral remote sensing image data of tea canopy dimensions, wherein the multispectral remote sensing image data comprises DN value data of a plurality of wave bands, and each wave band corresponds to one DN value image;

obtaining DN value data of a reference white board and known reflectivity data of the reference white board corresponding to each wave band;

and determining multispectral reflectivity data of the tea canopy scale by utilizing DN value data of the tea canopy scale of the same wave band, DN value data of the reference white board and reflectivity data of the wave band corresponding to the reference white board.

Alternatively, the spectral reflectance data for each band is calculated by:

R_i＝DN_i/(DN_w/R_w)

Wherein R _i is spectral reflectance data of an ith wave band, DN _i is DN value data of a tea canopy in the ith wave band, DN _w is DN value data of a reference white board, and R _w is known reflectance data of the reference white board corresponding to the ith wave band.

Optionally, the multispectral reflectance data of the tea canopy scale comprises a reflectance of a red side 1 wave band, a reflectance of a near infrared wave band, a reflectance of a red side 2 wave band, a reflectance of a green wave band, a reflectance of a blue wave band and a reflectance of a red wave band;

The constructing a multispectral-based full-channel difference vegetation index by utilizing multispectral reflectivity data of the tea canopy scale comprises the following steps:

according to the reflectivity of the red 1 wave band, the reflectivity of the near infrared wave band, the reflectivity of the red 2 wave band, the reflectivity of the green wave band, the reflectivity of the blue wave band and the reflectivity of the red wave band, constructing a multispectral-based total channel difference vegetation index with the following formula:

FCDVI=2*(ED1+NIR)+ED2+G-8*(B+R)

Wherein FCDVI is a full-channel difference vegetation index, ED1 is the reflectivity of the red 1 band, NIR is the reflectivity of the near infrared band, ED2 is the reflectivity of the red 2 band, G is the reflectivity of the green band, B is the reflectivity of the blue band, and R is the reflectivity of the red band.

Optionally, the constructing a spectrum index for detecting tea based on the full-channel difference vegetation index includes:

acquiring a normalized difference vegetation index, and constructing a spectrum index for detecting tea according to the following formula based on the full-channel difference vegetation index and the normalized difference vegetation index:

T1=(NDVI-FCDVI)/(NDVI+FCDVI)

Wherein TI is a spectral index for detecting tea, NDVI is a normalized difference vegetation index, FCDVI is a full-channel difference vegetation index.

Optionally, the method further comprises:

Constructing a plurality of indexes based on multispectral reflectivity data of tea canopy scales, wherein the indexes comprise normalized difference vegetation indexes, red edge normalized difference indexes, red edge chlorophyll indexes, MERIS land chlorophyll indexes, modified chlorophyll absorption ratio indexes, enhanced vegetation indexes, transformed chlorophyll absorption reflection indexes, first modified ratio, second modified ratio, green band chlorophyll indexes, green normalized difference vegetation indexes, plant senescence reflection indexes, canopy chlorophyll content indexes, nonlinear vegetation indexes and triangular greenness indexes;

Constructing a prediction model according to the indexes and the normalized difference vegetation indexes, and training the prediction model to obtain a trained model;

and predicting the chlorophyll content of the tea based on the trained model.

In a second aspect, the present invention also provides a spectral index construction device for detecting tea leaves, comprising:

The data acquisition module is used for acquiring multispectral reflectivity data of the tea canopy scale;

the first construction module is used for constructing a full-channel difference vegetation index based on multiple spectrums by utilizing the multispectral reflectivity data of the tea canopy scale;

And the second construction module is used for constructing a spectrum index for detecting tea based on the full-channel difference vegetation index.

In a third aspect, the invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of spectral index construction for detecting tea leaves as described in the first aspect above when executing the computer program.

In a fourth aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method of spectral index construction for detecting tea leaves as described in the first aspect above.

In a fifth aspect, the invention also provides a computer program product comprising a computer program which when executed by a processor implements a method of spectral index construction for detecting tea leaves as described in the first aspect above.

Based on the technical scheme, the invention has the following beneficial effects compared with the prior art:

According to the method and the device for constructing the spectral index for detecting the tea, the multispectral reflectivity data of the tea canopy scale are obtained, the data comprise the reflectivity information of different wavebands, the physiological parameters of the tea can be evaluated more accurately, and the growth condition and the nutrition state of the tea can be reflected more comprehensively. By utilizing the acquired multispectral reflectivity data to construct a multispectral-based full-channel difference vegetation index, the physiological parameter changes of different vertical structure positions of the tea canopy can be reflected more sensitively. This index improves the accuracy and precision of the assay and helps to find subtle differences in tea growth. The spectral index for detecting the tea is constructed based on the full-channel difference vegetation index, so that the spectral index can more accurately reflect the growth condition and the nutrition state of the tea, and the detection accuracy is improved. The spectrum index constructed by the method is designed according to specific physiological parameters (such as chlorophyll content, nitrogen content and the like) of the tea, and can be detected specifically according to the specific physiological parameters of the tea.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic flow chart of a method for constructing a spectrum index for detecting tea leaves.

Fig. 2 is a schematic diagram of a spectral image processing procedure provided by the present invention. Wherein, (a) is an original multispectral image after blade scale pretreatment, (b) is a multispectral image after blade scale enhancement by FCDVI, (c) is an image after blade scale binarization by Otsu, (d) is a multispectral image after masking the original multispectral image by the blade scale binarization image, (e) is an original multispectral image after canopy scale pretreatment, (f) is a multispectral image after canopy scale enhancement by FCDVI, (g) is an image after canopy scale binarization by Otsu, and (h) is a multispectral image after masking the original multispectral image by the canopy scale binarization image.

Fig. 3 is a graph of registration fusion effects provided by the present invention. Wherein, (a) is a RGB three-channel combined image before registration of the blade-scale multispectral image, (b) is a RGB three-channel combined true color image after registration of the blade-scale multispectral image, (c) is a RGB three-channel combined image before registration of the canopy-scale multispectral image, and (d) is a RGB three-channel combined true color image after registration of the canopy-scale multispectral image.

Fig. 4 is a schematic structural diagram of a spectrum index constructing device for detecting tea leaves.

Fig. 5 is a schematic structural diagram of an electronic device provided by the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The spectral index construction method and apparatus for detecting tea leaves of the present invention are described below with reference to fig. 1-4.

Referring to fig. 1, the method for constructing the spectrum index for detecting tea leaves comprises the following steps:

and S110, acquiring multispectral reflectivity data of the tea canopy scale.

The goal of this step is to collect reflectance data for the tea canopy at different spectral bands. These data will be used for subsequent spectral index construction and analysis. Multispectral remote sensing instruments, such as multispectral cameras or spectrometers, may be used that are capable of acquiring reflectance data in multiple bands simultaneously. The weather is clear and the cloud shielding is avoided during measurement, so that the influence of atmospheric conditions on a measurement result is reduced. At the same time, the measurement should be taken at a height above the tea canopy to ensure that the data obtained is representative of the average reflectance of the whole canopy. The reflectance data for each band is recorded and saved in a digital format for subsequent processing and analysis.

And step S120, constructing a full-channel difference vegetation index based on multiple spectrums by utilizing the multispectral reflectivity data of the tea canopy scale.

And constructing a full-channel difference vegetation index capable of reflecting the growth condition and the nutrition state of the tea by using the acquired multispectral reflectivity data. From the acquired multispectral data, a wavelength band which is sensitive to the growth of tea and can reflect the nutrition state of the tea is selected. These bands may include the red 1 band (ED 1), the near infrared band (NIR), the red 2 band (ED 2), the green band (G), the blue band (B), and the red band (R), each of which play a critical role in vegetation monitoring. And constructing a full-channel difference vegetation index by adopting modes of summation, difference or weighted summation and the like through the spectral reflectivity data of the wave bands. The index integrates information of a plurality of wave bands, can comprehensively reflect the reflection characteristic difference of the tea canopy under different wave bands, and avoids the information loss possibly caused by a single wave band, thereby reflecting the reflection characteristic of the tea canopy and the relation between the reflection characteristic and the growth condition of the tea canopy more comprehensively. The calculated FCDVI value can be used as an important index for monitoring the growth of the tea, and is used for evaluating the growth state, the nutrition level and the possible growth stress of the tea.

And step S130, constructing a spectrum index for detecting tea based on the full-channel difference vegetation index.

In the process of tea garden management and tea quality assessment, accurate spectrum index is a key for measuring the growth condition and nutrition level of tea. In step S130, a spectral index dedicated to detecting tea leaves is further constructed on the basis of the full-channel difference vegetation index. The index may be constructed based on the full path difference vegetation index FCDVI described above, as well as other vegetation indices, such as Normalized Difference Vegetation Index (NDVI), chlorophyll vegetation index (Chlorophyll Vegetation Index, CVI), soil-adjusted vegetation index (Soil-Adjusted Vegetation Index, SAVI), enhanced vegetation index (Enhanced Vegetation Index, EVI), red edge location index (Red Edge Position Index, REPI), normalized water index (Normalized Water Index, NWI), and the like.

The above CVI evaluates the chlorophyll content by calculating the difference in reflectivity of red and near infrared bands using the specific absorption characteristics of chlorophyll in these two bands. For tea, CVI can reflect the relative content of chlorophyll in tea leaves, and further indirectly indicate the growth state and nutrition condition of the tea. The SAVI is used for correcting the influence of soil background on the basis of NDVI, and is particularly suitable for environments with large change of soil background. In the tea planting area, SAVI can provide more accurate vegetation information if the soil background has a significant impact on spectral reflectance. The EVI is an improvement on NDVI, and the evaluation accuracy of vegetation coverage and biomass is improved by introducing an atmospheric correction factor and a soil adjustment coefficient. In tea garden management, EVI can more effectively monitor the growth dynamics and health status of tea. The red edge position refers to a steep edge position of transition from a red light region to a near infrared region in a chlorophyll absorption spectrum, and the REPI evaluates chlorophyll content and health status of plants by measuring changes of the position. For tea, REPI can reflect the slight change of chlorophyll in tea leaves, and is helpful for timely finding abnormal growth. The NWI is mainly used for monitoring water and soil moisture, but can also be used as an auxiliary index for evaluating the moisture condition and irrigation requirement of tea in tea garden management. NWI helps to guide water management in tea gardens by monitoring the moisture content of tea leaves.

The indexes can reflect the growth condition and nutrition level of the tea based on the reflection characteristics of the tea under different spectral bands. Because the application scenes and the sensitivity of different spectrum indexes can be different, a person skilled in the art can comprehensively consider according to specific monitoring requirements and tea garden environments, and select one or more indexes to construct the spectrum indexes for detecting the tea.

According to the method for constructing the spectral index for detecting the tea, the multispectral reflectivity data of the tea canopy scale are obtained, the data comprise reflectivity information of different wavebands, the physiological parameters of the tea can be evaluated more accurately, and the growth condition and the nutrition state of the tea can be reflected more comprehensively. By utilizing the acquired multispectral reflectivity data to construct a multispectral-based full-channel difference vegetation index, the physiological parameter changes of different vertical structure positions of the tea canopy can be reflected more sensitively. This index improves the accuracy and precision of the assay and helps to find subtle differences in tea growth. The spectral index for detecting the tea is constructed based on the full-channel difference vegetation index, so that the spectral index can more accurately reflect the growth condition and the nutrition state of the tea, and the detection accuracy is improved. The spectrum index constructed by the method is designed according to specific physiological parameters (such as chlorophyll content, nitrogen content and the like) of the tea, and can be detected specifically according to the specific physiological parameters of the tea.

According to the spectral index construction method for detecting tea provided by the invention, multispectral reflectivity data of tea canopy dimensions are obtained through a remote sensing technology or a special spectrometer. The process is relatively simple and rapid, and complicated sampling, processing and detecting steps in the traditional laboratory physicochemical detection are avoided, so that the time and labor cost are greatly saved. After multispectral reflectivity data are acquired, the method utilizes a computer to perform automatic processing and analysis, and the detection efficiency is further improved.

According to the spectral index construction method for detecting tea provided by the invention, the reflectance data of a plurality of points of the tea canopy can be obtained simultaneously by adopting a surface scale measurement mode through a multispectral remote sensing instrument, so that the problem of low efficiency of point location measurement is avoided. Through surface scale measurement, the method can evaluate the growth condition and the nutrition state of the tea more comprehensively, and provides more accurate information support for the establishment of cultivation management measures in tea gardens. The method adopts a spectrum nondestructive detection technology, does not cause any damage to the tea sample, and realizes real-time monitoring of the growth condition of the tea. By rapidly acquiring and processing multispectral reflectivity data, the method can reflect the growth condition and the nutrition state of tea in time, and provides timely decision support for tea garden management.

In an alternative embodiment, the acquiring the spectral reflectance data of the canopy dimensions of the tea leaf in step S110 includes:

S1101, acquiring multispectral remote sensing image data of tea canopy dimensions, wherein the multispectral remote sensing image data comprises DN value data of a plurality of wave bands, and each wave band corresponds to one DN value image.

This step is to acquire remote sensing image data of the tea canopy in different spectral bands, which are usually represented in Digital form, i.e. as DN value (Digital Number) data. The multispectral remote sensing instrument can simultaneously capture images of a plurality of wave bands, and each wave band corresponds to one DN value image. The bands cover visible light, near infrared and other areas, and can reflect the reflection characteristics of the tea canopy in different spectral ranges.

S1102, DN value data of a reference white board and known reflectivity data of the reference white board corresponding to each wave band are acquired.

The reference whiteboard is a standard object for calibrating the remote sensing instrument, the reflectivity of which is known in various bands. Before or after multispectral remote sensing image data of tea canopy dimensions are acquired, DN value data of a reference whiteboard is required to be measured. At the same time, it is also necessary to obtain known reflectivity data of the reference whiteboard at each band, which is provided by the manufacturer or is strictly calibrated.

S1103, determining multispectral reflectivity data of tea canopy dimensions by using DN value data of tea canopy dimensions, DN value data of a reference white board and reflectivity data of a corresponding wave band of the reference white board in the same wave band.

For each band, the reflectance of the tea canopy can be calculated using the following equation:

tea canopy reflectance= (tea canopy DN value-dark current DN value)/(reference whiteboard DN value-dark current DN value) ×reference whiteboard reflectance

The dark current DN value refers to the DN value output by the instrument under the condition of no illumination, and is used for calibrating and removing noise inside the instrument. In actual operation, since the dark current DN value is usually small and stable, it can be ignored, simplifying the calculation process. Thus, the above formula can be reduced to:

R_i＝DN_i/(DN_w/R_w)

wherein, R _i is spectral reflectance data of the ith wave band, DN _i is DN value data of the tea canopy in the ith wave band, DN _w is DN value data of the reference white board, and R _w is known reflectance data of the reference white board corresponding to the ith wave band (namely, standard reflectance of the reference white board in the wave band).

By this step, reflectance data for the tea canopy at each band can be obtained and used for subsequent spectral index construction and analysis.

When the remote sensing instrument collects data, the remote sensing instrument may be affected by internal noise, optical distortion and other factors, so that errors occur in the data. External environmental factors (e.g., light intensity, atmospheric conditions, etc.) may also affect the data as it is collected. According to the invention, the white board is referenced for calibration, so that the influence of the internal noise and external environmental factors of the instrument on the data can be eliminated, and the accuracy of the tea canopy scale spectral reflectance data is improved. The same reference white board is used for calibration, so that consistency and comparability of data collected at different times and different places can be ensured, and subsequent data analysis and application are facilitated.

In an alternative embodiment, the invention uses an MS 600V 2 multispectral camera to collect multispectral remote sensing image data. The camera has multiple channels including Blue (B), green (G), red (R), red Edge (ED 1), RED EDGE LP (ED 2), and NEAR INFRARED (NIR). Because each channel sensor is distributed in an array, the acquired images of each channel may have a deviation in spatial position. However, the camera is not equipped with an automatic registration procedure, and the multispectral remote sensing image data is a multispectral image in order to facilitate band fusion and spectrum information sampling. Firstly, extracting an average DN value of a whiteboard in each image from the multispectral image, converting the multispectral DN value image into a reflectivity image in python 3.6 according to a formula 1 by combining the reflectivity of each wave band calibrated by a reference whiteboard, automatically selecting and matching the characteristics of each wave band image by using a Sift algorithm, and then fusing wave bands r.

Specifically, the multispectral image is preprocessed. This includes finding and extracting the average DN (Digital Number) values of the white board in each image in the acquired multispectral image. The DN value is the luminance value of the image pixel and is used to represent the luminance information of the image. The whiteboard is used for calibration and standardization because its reflectivity is close to 100% and therefore can be used as a reference for reflectivity calculation. The multispectral DN value image is converted to a reflectance image using equation 1 in combination with the reflectance of each band calibrated with the reference whiteboard. The converted reflectivity image is more convenient for subsequent analysis and processing of spectral information.

To solve the problem of the deviation of the images of each channel in the spatial position, a Sift (Scale-INVARIANT FEATURE TRANSFORM) algorithm is used for image registration. The Sift algorithm is an algorithm for detecting and describing local features of an image, and can identify stable feature points under different scale, rotation, illumination and other conditions. And automatically selecting characteristic points in the images of each wave band by using a Sift algorithm, and carrying out characteristic matching. Feature matching refers to finding the same feature points in different images, thereby determining the relative positional relationship between the images. And (3) carrying out wave band fusion on the basis of feature matching. Band fusion refers to combining images of multiple bands into one image according to a certain rule so as to sample and analyze the subsequent spectrum information. Such as aligning and combining the pixel values of the respective channel images according to their spatial locations. Finally, the effect before and after registration fusion is displayed. By comparing the images before and after registration, the effects of image registration and fusion can be intuitively seen, and whether the consistency of the images of each channel in spatial positions is improved.

The effect of registering the fusion before and after fusion is shown in figure 3. Wherein, (a) is a RGB three-channel combined image before registration of the blade-scale multispectral image, (b) is a RGB three-channel combined true color image after registration of the blade-scale multispectral image, (c) is a RGB three-channel combined image before registration of the canopy-scale multispectral image, and (d) is a RGB three-channel combined true color image after registration of the canopy-scale multispectral image.

Before fusion, the images of each spectrum band may have problems of space deviation, uneven brightness, noise pollution and the like due to factors such as sensor array distribution, atmospheric interference, instrument noise and the like. Through fusion calibration, the spatial deviation of each band image is corrected, the brightness is uniformly adjusted, the noise is suppressed, and the overall image quality is remarkably improved. The images are clearer and finer, and the subsequent analysis and processing are easier. In addition, before fusion, due to factors such as sensor array distribution, camera internal processing and the like, the images of each wave band may have deviation in spatial position, so that the phenomenon of dislocation or overlapping occurs during image splicing. Through feature matching and wave band fusion technology, the spatial positions of the images of all wave bands are accurately aligned, and the image stitching is more natural and accurate. The spatial consistency is remarkably improved, and a reliable basis is provided for subsequent spectrum information sampling and analysis.

In an alternative embodiment, in order to effectively extract a partial discrete spectrum curve of the canopy and avoid the influence of the background such as shadows in the image, the invention proposes a Full-channel difference vegetation index (Full CHANNEL DIFFERENCE vegetation index, FCDVI) based on a multispectral image by performing spectral feature analysis and multiple experiments on the canopy and other features. The multispectral reflectance data of the tea canopy scale comprises the reflectance of the red 1 band, the reflectance of the near infrared band, the reflectance of the red 2 band, the reflectance of the green band, the reflectance of the blue band and the reflectance of the red band, as shown in table 1.

TABLE 1

The constructing a multispectral-based full-channel difference vegetation index according to the step S120 by using the multispectral reflectance data of the tea canopy scale includes:

s1201, constructing a multispectral-based total channel difference vegetation index according to the reflectivity of a red side 1 wave band, the reflectivity of a near infrared wave band, the reflectivity of a red side 2 wave band, the reflectivity of a green wave band, the reflectivity of a blue wave band and the reflectivity of the red wave band:

FCDVI = 2 x (ed1+nir) +ed2+g-8 x (b+r), equation 2;

Wherein FCDVI is a full-channel difference vegetation index, ED1 is the reflectivity of the red 1 band, NIR is the reflectivity of the near infrared band, ED2 is the reflectivity of the red 2 band, G is the reflectivity of the green band, B is the reflectivity of the blue band, and R is the reflectivity of the red band. The index can greatly enhance the tea characteristics in the multispectral image for both the leaf scale and canopy scale multispectral image, and increase the contrast with other ground objects so as to distinguish tea from the background.

The full-channel difference vegetation index FCDVI combines the reflectivity data of a plurality of spectral bands, and can reflect the vegetation characteristics of the tea canopy more comprehensively. FCDVI has higher accuracy and precision in vegetation monitoring compared to the vegetation index of a single spectral band. Through a specific mathematical model FCDVI, the vegetation information of the tea canopy can be enhanced, and the influence of background information is weakened. This helps to accurately identify the tea canopy in a complex environment, providing reliable data support for subsequent remote sensing applications. FCDVI takes data of a plurality of spectral bands into consideration in the construction process, so that the method has strong adaptability. The method can be used under different illumination conditions, observation angles and earth surface coverage types, and provides a wider application scene for remote sensing monitoring of tea canopy.

A schematic diagram of the spectral image processing procedure is shown in fig. 2. Wherein, (a) is an original multispectral image after blade scale pretreatment, (b) is a multispectral image after blade scale enhancement by FCDVI, (c) is an image after blade scale binarization by Otsu, (d) is a multispectral image after masking the original multispectral image by the blade scale binarization image, (e) is an original multispectral image after canopy scale pretreatment, (f) is a multispectral image after canopy scale enhancement by FCDVI, (g) is an image after canopy scale binarization by Otsu, and (h) is a multispectral image after masking the original multispectral image by the canopy scale binarization image.

First, analysis is made of (a) to (d) in fig. 2, where (a) is the original multispectral image after blade-scale preprocessing, including noise, background interference, and reflection differences between different blades. The blade scale uses FCDVI enhanced multispectral image (b), so that noise and background interference are reduced, and blade characteristics are more prominent. The blade scale is converted into a binary image containing only two colors of black and white using Otsu binarized image (c). The leaves are identified as white (foreground) and the background and other non-leaf portions are identified as black (background), the binarized image simplifying the image analysis. The binarized image is used as a mask to extract the leaf area in the original multispectral image, and a multispectral image (d) after the original multispectral image is masked by the leaf scale with the binarized image is obtained, and the masked image only contains information of the leaf area, so that the interference of the background and other non-leaf parts is removed, and the spectral characteristics of the leaf can be analyzed more accurately.

Next, analysis is performed on fig. 2 (e) through (h), which are raw multispectral images after canopy-scale preprocessing, similar to raw images on the blade scale. It contains reflection or transmission information of the whole canopy in different spectral bands. Similar to the leaf scale, FCDVI is used to enhance the vegetation characteristics of the canopy scale, resulting in a multispectral image (f) of the canopy scale enhanced with FCDVI, which may more clearly show the chlorophyll content, health status, canopy structure, etc. characteristics of the canopy. At the canopy scale, otsu binarization is used to convert the image into a binary image to distinguish the canopy from the background, resulting in an image (g) of the canopy scale binarized using Otsu. At the canopy scale, the binarized image is used as a mask to extract the canopy region in the original multispectral image, resulting in a multispectral image (h) of the canopy scale after masking the original multispectral image with the binarized image. The image after masking contains only the information of the canopy region, removing the interference of the background and other non-canopy parts. This helps to more accurately analyze the spectral characteristics of the canopy and reduces the effects of background noise.

The multispectral image enhanced by FCDVI shows vegetation features more clearly on the blade scale and canopy scale, which is helpful for reducing noise and background interference. Otsu binarization converts the image into a binary image, simplifying image analysis. The use of the binarized image as a mask allows for the extraction of the blade or canopy region of interest, thereby more accurately analyzing the spectral features. The interference of the background and other non-interested areas is removed from the multispectral image after masking, so that the subsequent analysis is more accurate and reliable, and the accuracy and efficiency of the image analysis are improved.

The original multispectral image and two groups of canopy scale multispectral images enhanced by FCDVI are subjected to the spectrum index calculation formula (namely formula 3) for tea detection, which is applied to the enhanced multispectral image. Pairing the calculated vegetation index with physiological parameters (such as nitrogen content, chlorophyll content and the like) of the tea to form a data set. Statistical methods (e.g., pearson correlation coefficient, spearman rank correlation coefficient, etc.) are used to evaluate the correlation between the vegetation index and the physiological parameter. And comparing the correlation between the vegetation index obtained by the original image and the FCDVI enhanced image and the physiological parameter, and analyzing the specific effect of FCDVI treatment on the improvement of the correlation.

Through comparative analysis, the multispectral image processed by FCDVI is found to show remarkable advantages in the correlation of vegetation indexes and physiological parameters of tea, and is particularly embodied in the increment of number and the improvement of correlation coefficients of remarkably correlated data. This demonstrates that FCDVI proposed by the present invention for MS 600V 2 type multispectral sensor can significantly enhance the tea characteristics, greatly helping to reject soil and shadow noise.

In an alternative embodiment, to more accurately detect the growth status and health of tea, the present invention constructs a new spectral index (TI) for detecting tea based on the full path difference vegetation index (FCDVI) and the Normalized Difference Vegetation Index (NDVI). The index combines the advantages of two vegetation indexes, and aims to improve the accuracy and reliability of tea detection. The constructing a spectrum index for detecting tea based on the full-channel difference vegetation index in the step S130 includes:

s1301, acquiring a normalized difference vegetation index, and constructing a spectrum index for detecting tea according to the following formula based on the full-channel difference vegetation index and the normalized difference vegetation index:

t1= (NDVI-FCDVI)/(ndvi+ FCDVI), equation 3;

The Normalized Difference Vegetation Index (NDVI) is one of important parameters reflecting crop growth vigor and nutrition information, and its calculation formula is ndvi= (NIR-R)/(nir+r), where NIR is the reflectivity of the near infrared band and R is the reflectivity of the red band. And obtaining the reflectivity data of the near infrared band and the red band of the tea canopy by remote sensing image or ground spectrum measuring equipment, and substituting the reflectivity data into an NDVI formula to calculate so as to obtain an NDVI value.

The full path difference vegetation index (FCDVI) is obtained by combining reflectance data of multiple spectral bands of the tea canopy by a mathematical model as in equation 2. Similarly, reflectivity data of the wavebands are obtained through remote sensing images or ground spectrum measuring equipment, and then the reflectivity data are substituted into a FCDVI formula to calculate, so that FCDVI values are obtained.

Based on the obtained NDVI and FCDVI values, a spectral index (TI) for detecting tea was constructed. The calculation formula is the formula 3. By this formula, the difference and sum of NDVI and FCDVI can be combined to form a new spectral index TI reflecting the specific vegetation characteristics of the tea canopy.

NDVI and FCDVI have different emphasis and sensitivity in reflecting vegetation characteristics. NDVI mainly reflects the coverage and growth conditions of vegetation, and FCDVI combines information of multiple spectral bands, and can reflect the vegetation characteristics of the tea canopy more carefully. The two are combined to form complementation, so that the growth state and the health condition of the tea can be reflected more comprehensively. By constructing the TI spectral index, the advantages of NDVI and FCDVI can be fused together. TI can reflect the vegetation characteristics of the tea canopy more accurately, and improve the accuracy and reliability of tea detection. In the remote sensing monitoring of tea, TI can more accurately identify tea canopy and extract key information related to tea growth. This helps to reduce misjudgment and missed judgment, and improves the accuracy of tea identification.

According to the invention, by combining the advantages of the NDVI and FCDVI vegetation indexes, TI can reflect the vegetation characteristics of the tea canopy more comprehensively, so that the accuracy of tea detection is improved. NDVI is sensitive to vegetation coverage and growth conditions, whereas FCDVI combines information in multiple spectral bands, which can enhance vegetation information and attenuate the impact of background information. Therefore, TI can more accurately reflect the growth state and health condition of tea. The construction process of TI considers data of a plurality of spectrum bands, so that the method has strong information extraction capability. The method can accurately identify the tea canopy in a complex environment, extract key information related to tea growth, and provide reliable data support for subsequent remote sensing application. By monitoring TI change of the tea canopy, information such as growth condition, health condition and yield potential of tea can be known in time. The method is helpful for guiding the production and management of tea, improving the quality and yield of the tea and providing powerful support for the sustainable development of the tea industry.

The method ensures that the selected wave band can sensitively reflect the growth condition and the nutrition state of the tea by selecting and optimizing the wave band of the multispectral data. These bands generally cover the visible, near infrared, and short wave infrared regions, and capture the reflection characteristics of the tea canopy in different spectral ranges. After the multispectral reflectivity data is acquired, the method performs data preprocessing, including removing noise, correcting instrument errors and the like, so as to ensure the accuracy and reliability of the data. This step helps to reduce errors due to instrumentation or environmental factors, providing a high quality data base for subsequent analysis. The method adopts a canopy dimension measurement mode, and obtains multispectral reflectivity data of the whole tea canopy by a remote sensing technology or a special spectrometer. This approach can reduce interference of the background environment (such as soil, weeds, etc.) on the extraction of the tea information, because the canopy reflectivity mainly reflects the growth condition of the tea itself. The method can further weaken the influence of the background environment by constructing a full-channel difference vegetation index based on multiple spectrums. The vegetation index is a comprehensive index reflecting vegetation growth conditions, and the characteristic information of the tea canopy can be highlighted by calculating the difference of reflectivities of different wave bands, so that the interference of a background environment is reduced. The spectral index constructed by the method is designed for specific physiological parameters (such as chlorophyll content, nitrogen content and the like) of the tea. The targeted design enables the spectrum index to reflect the growth condition and the nutrition state of the tea more accurately, and improves the detection accuracy.

In an alternative embodiment, a plurality of vegetation indices may be constructed based on the multispectral reflectance data and used in conjunction with a Normalized Difference Vegetation Index (NDVI) to construct a predictive model to predict leaf chlorophyll content. The invention selects a plurality of vegetation indexes and original wave bands as variables, and the variables are used for subsequent model training. Although the vegetation index dimension is low, the machine learning algorithm is chosen to handle redundancy issues such as co-linearity among the variables, so all selected variables are incorporated into the model.

The method further comprises the steps of:

S210, constructing a plurality of indexes based on multispectral reflectivity data of tea canopy scales, wherein the indexes comprise normalized difference vegetation indexes, red side normalized difference indexes, red side chlorophyll indexes, MERIS land chlorophyll indexes, corrected chlorophyll absorption ratio indexes, enhanced vegetation indexes, transformed chlorophyll absorption reflection indexes, first corrected ratio, second corrected ratio, green band chlorophyll indexes, green normalized difference vegetation indexes, plant senescence reflection indexes, canopy chlorophyll content indexes, nonlinear vegetation indexes and triangular greenness indexes. The formulas of these indices are shown in table 2. ED in Table 2 refers to the reflectance in the red band and NIR2 refers to the reflectance in the near infrared band. The meaning of the other parameters is consistent with that in the previous description.

TABLE 2

Multispectral reflectance data for the canopy scale of the tea was collected. Such data is typically acquired by remote sensing devices or terrestrial spectral measuring devices, covering information in different spectral bands. Based on the collected multispectral reflectance data, a plurality of vegetation indices are constructed. These indices include Normalized difference vegetation Index (Normalized Difference Vegetation Index, NDVI), red-edge Normalized difference Index (Normalized DIFFERENCE RED EDGE, NDRE), red-edge chlorophyll Index (Chlorophyll Index Red Edge, CIRE), MERIS land chlorophyll Index (MERIS TERRESTRIAL Chlorophyll Index, MTCI), modified chlorophyll absorption ratio Index (Modified Chlorophyll Absorption Ratio Index, MCARI), enhanced vegetation Index (Enhanced Vegetation Index, EVI), transformed chlorophyll absorption reflection Index (Transformed CARI, TCARI), first modified ratio (Modified Simple Ratio, msr 1), second modified ratio (Modified Simple Ratio, msr 2), green-band chlorophyll Index (Chlorophyll Index Green, CIG), green Normalized difference vegetation Index (Green Normalized Difference Vegetation Index, GNDVI), plant senescence reflection Index (PLANT SENESCENCE REFLECTANCE Index, PSRI), canopy chlorophyll content Index (Canopy Chlorophyll Content Index, CCCI), nonlinear vegetation Index (Nonlinear Vegetation Index, NLVI), triangular greenness Index (Triangular Greenness Index, TGI), and the like. MERIS is a medium resolution imaging spectrometer and CARI is chlorophyll absorption reflectance index. These indices can reflect the vegetation characteristics of the tea canopy, such as chlorophyll content, vegetation coverage, growth status, etc.

S220, constructing a prediction model according to the indexes and the normalized difference vegetation indexes, and training the prediction model to obtain a trained model.

And jointly constructing a prediction model by taking the constructed multiple vegetation indexes and the NDVI as input variables. The model is trained by selecting a suitable machine learning algorithm, such as a Support Vector Machine (SVM), partial Least Squares (PLS), neural network, etc. Through the training process, parameters and structures of the model are optimized, so that chlorophyll content of tea can be accurately predicted.

The Random Forest (RF) is an integrated learning method, and the accuracy and stability of the model are improved by constructing a plurality of decision trees and synthesizing the prediction results thereof. Each decision tree is trained based on a randomly selected pilot sample and input variables. Partial least squares regression (PLS) is a multivariate statistical data analysis method that is particularly useful when there are multiple collinearity between process variables or when the number of variables is greater than the number of samples. The dominant factor with the strongest interpretation power on the dependent variable is extracted by combining the advantages of principal component analysis, canonical analysis and linear regression. A Support Vector Machine (SVM) is a kernel-based machine learning algorithm that is capable of mapping input variables into a high-dimensional feature space, thereby processing high-dimensional input vectors. SVM is widely used in the fields of spectrum analysis and the like, and produces accurate calibration results. The BP neural network is a multi-layer feedforward neural network and is trained through an error back propagation algorithm. It comprises an input layer, a hidden layer and an output layer, minimizing prediction errors by adjusting network parameters.

S230, predicting chlorophyll content of the tea based on the trained model.

And verifying the trained model by using an independent verification data set, and evaluating the prediction performance and accuracy of the model. If the model performance is good, the model can be applied to actual tea production, and is used for predicting the chlorophyll content of tea and guiding the planting and management of the tea.

According to the method, the plurality of vegetation indexes are constructed, and the prediction model is jointly constructed by combining with the NDVI, so that the information in multispectral reflectivity data can be fully utilized, and the accuracy of predicting the chlorophyll content of the tea leaves is improved. Different vegetation indexes can reflect different characteristics of tea canopy, and the combination of the different vegetation indexes can form complementation, so that the prediction capability of the model is improved. The multiple vegetation indexes are used as input variables, so that the diversity and complexity of the model can be increased, and the model can be better adapted to the tea growth conditions under different environments and conditions. This helps to enhance the robustness of the model so that it maintains stable predictive performance in tea production at different times and locations. By predicting the chlorophyll content of the tea, the growth condition and the health condition of the tea can be known in time, and powerful support is provided for production management of the tea. The method fully utilizes the advantages of the remote sensing technology, constructs a plurality of vegetation indexes through multispectral reflectivity data, and is successfully applied to the prediction of the chlorophyll content of the tea leaves.

In order to evaluate the performance of the model, the invention adopts a leave-one-out cross-validation method. This method divides the data set into n subsets, leaving one subset at a time as the test set and the remaining subsets as the training set for model training. And then, predicting the test set by using the trained model, and calculating the accuracy of the prediction result. This process is repeated n times, with each subset being used as a test set.

For quantifying the accuracy of the model, the invention calculates two indexes of the goodness of fit R2 and the root mean square error RMSE, and the calculation formula can refer to the description in the prior art. The closer R2 is to 1, the better the fitting effect of the model is, the smaller the RMSE is, the smaller the error of the prediction result is, and the higher the precision is.

According to the invention, a prediction model is constructed by selecting a plurality of machine learning algorithms and vegetation indexes, and the performance of the model is verified by a left-right cross-verification method, an R2 and RMSE and other precision evaluation methods. The method provides an effective solution for vegetation index and modeling, and helps to improve accuracy and stability of the model. The person skilled in the art may also choose to use a plurality of indices as described above to construct a predictive model of nitrogen content or other factors related to tea growth.

The spectral index constructing apparatus for detecting tea leaves provided by the present invention will be described below, and the spectral index constructing apparatus for detecting tea leaves described below and the spectral index constructing method for detecting tea leaves described above may be referred to correspondingly to each other.

The spectral index constructing apparatus for detecting tea leaves provided by the invention, referring to fig. 4, comprises:

A data acquisition module 310, configured to acquire multispectral reflectance data of a tea canopy scale;

A first construction module 320, configured to construct a multispectral-based full-channel difference vegetation index using multispectral reflectance data of the tea canopy scale;

A second construction module 330 is configured to construct a spectral index for detecting tea based on the full path difference vegetation index.

Fig. 5 illustrates a physical schematic diagram of an electronic device, which may include a processor (processor) 410, a communication interface (Communications Interface) 420, a memory (memory) 430, and a communication bus 440, as shown in fig. 5, where the processor 410, the communication interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 may invoke logic instructions in the memory 430 to perform a spectral index construction method for detecting tea leaves.

Further, the logic instructions in the memory 430 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. The storage medium includes a U disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, an optical disk, or other various media capable of storing program codes.

In another aspect, the present invention also provides a computer program product comprising a computer program storable on a non-transitory computer readable storage medium, the computer program, when executed by a processor, being capable of performing the method of constructing a spectral index for detecting tea leaves provided by the methods described above.

In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the spectral index construction method for detecting tea leaves provided by the methods described above.

The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.

It should be noted that the above-mentioned embodiments are merely for illustrating the technical solution of the present invention, and not for limiting the same, and although the present invention has been described in detail with reference to the above-mentioned embodiments, it should be understood by those skilled in the art that the technical solution described in the above-mentioned embodiments may be modified or some technical features may be equivalently replaced, and these modifications or substitutions do not make the essence of the corresponding technical solution deviate from the spirit and scope of the technical solution of the embodiments of the present invention.

Claims

1. A method of spectral index construction for detecting tea leaves, comprising:

Obtaining multispectral reflectivity data of tea canopy dimensions;

2. A method of constructing a spectral index for detecting tea leaves as claimed in claim 1 wherein said obtaining spectral reflectance data of the canopy dimensions of tea leaves comprises:

3. A method of constructing a spectral index for detecting tea leaves according to claim 2, wherein the spectral reflectance data for each band is calculated by the formula:

R_i=DN_i/(DN_w/R_w)

4. The method for detecting a spectral index configuration of tea leaves according to claim 1, wherein the multispectral reflectance data of the tea leaf canopy scale includes reflectance of red 1 band, reflectance of near infrared band, reflectance of red 2 band, reflectance of green band, reflectance of blue band and reflectance of red band;

FCDVI=2*(ED1+NIR)+ED2+G-8*(B+R)

Wherein FCDVI is the full-channel difference vegetation index, The reflectance of the red 1 band is NIR, the reflectance of the near infrared band, ED2 is the reflectance of the red 2 band, G is the reflectance of the green band, B is the reflectance of the blue band, and R is the reflectance of the red band.

5. The method of claim 1, wherein constructing the spectral index for detecting tea based on the full path difference vegetation index comprises:

T1=(NDVI-FCDVI)/(NDVI+FCDVI)

6. A method of spectral index construction for detecting tea leaves according to claim 1, further comprising:

and predicting the chlorophyll content of the tea based on the trained model.

7. A spectral index configuring apparatus for detecting tea leaves, comprising:

8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor, when executing the computer program, implements a method for constructing a spectral index for detecting tea leaves as claimed in any one of claims 1 to 6.

9. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements a spectral index construction method for detecting tea leaves as claimed in any one of claims 1 to 6.

10. A computer program product comprising a computer program which, when executed by a processor, implements a method of spectral index construction for detecting tea leaves as claimed in any one of claims 1 to 6.