CN119445650A - Virtual maintenance training assessment management method - Google Patents
Virtual maintenance training assessment management method Download PDFInfo
- Publication number
- CN119445650A CN119445650A CN202411390168.8A CN202411390168A CN119445650A CN 119445650 A CN119445650 A CN 119445650A CN 202411390168 A CN202411390168 A CN 202411390168A CN 119445650 A CN119445650 A CN 119445650A
- Authority
- CN
- China
- Prior art keywords
- virtual maintenance
- feature
- virtual
- feature map
- key frame
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
-
- 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
- G06V10/62—Extraction of image or video features relating to a temporal dimension, e.g. time-based feature extraction; Pattern tracking
-
- 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
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- 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
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
- G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
- G09B7/00—Electrically-operated teaching apparatus or devices working with questions and answers
-
- G—PHYSICS
- G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
- G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
- G09B9/00—Simulators for teaching or training purposes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Educational Administration (AREA)
- Educational Technology (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a virtual maintenance training assessment management method, which is characterized in that a camera is used for collecting a virtual maintenance monitoring video of an assessed person, and a data processing and image analysis algorithm based on artificial intelligence and machine vision is introduced into the rear end to analyze the virtual maintenance monitoring video, so that the relation between the semantic and time sequence relevance of the virtual maintenance action of the assessed person is captured, irregular operation or error action of the maintenance action is identified, and when abnormal operation or error action is detected, prompt information is sent out and an assessment report is generated. Therefore, more intelligent virtual maintenance training assessment management can be realized, and the maintenance action time sequence semantics of the assessment personnel are understood by performing overall analysis based on time sequence context on the action semantics of the assessment personnel, so that abnormal identification and error detection of maintenance training actions are performed, the application range and reliability of a virtual maintenance training assessment scheme are improved, and more objective and comprehensive assessment results are provided.
Description
Technical Field
The application relates to the field of intelligent management, in particular to a virtual maintenance training assessment management method.
Background
The vehicle maintenance training assessment is an important link for carrying out operation skill and maintenance capability assessment on maintenance personnel. Traditional maintenance training tests are mainly performed through actual vehicles or equipment to conduct operation exercises, and the capability of maintenance personnel in terms of fault handling, repair, tool use and the like is evaluated. However, this conventional method has problems such as susceptibility to human subjective factors, high cost, poor flexibility and safety, and the like.
The virtual technology allows maintenance personnel to train without physical equipment by simulating a real-world maintenance environment, so that the training cost can be reduced, risks possibly occurring in actual operation can be avoided, and the flexibility of maintenance training assessment is improved. However, in the conventional virtual maintenance training assessment management scheme, the assessment of the vehicle maintenance training assessment is generally performed by means of a standardized mode and a single rule, and the standardized assessment method is convenient for assessment and comparison, but can limit creativity and flexibility of maintenance personnel. That is, in actual maintenance, the same fault may have multiple reasons and investigation methods, and conventional virtual training assessment often has difficulty in simulating such diversity, which may result in too single training content and operation mode.
Disclosure of Invention
The present application has been made to solve the above-mentioned technical problems. The embodiment of the application provides a virtual maintenance training assessment management method, which is characterized in that a camera is used for collecting a virtual maintenance monitoring video of an assessed person, and a data processing and image analysis algorithm based on artificial intelligence and machine vision is introduced into the rear end to analyze the virtual maintenance monitoring video, so that the relation between the semantic meaning and time sequence relevance of the virtual maintenance action of the assessed person is captured, irregular operation or error action of the maintenance action is identified, and when abnormal operation or error action is detected, prompt information is sent out and an assessment report is generated. Therefore, more intelligent virtual maintenance training assessment management can be realized, and the maintenance action time sequence semantics of the assessment personnel are understood by performing overall analysis based on time sequence context on the action semantics of the assessment personnel, so that abnormal identification and error detection of maintenance training actions are performed, the application range and reliability of a virtual maintenance training assessment scheme are improved, and more objective and comprehensive assessment results are provided.
According to one aspect of the present application, there is provided a virtual maintenance training assessment management method, including:
Obtaining a virtual maintenance monitoring video of an inspected person;
performing key frame sampling on the virtual maintenance monitoring video to obtain a time sequence of virtual maintenance monitoring key frames;
performing multiscale feature extraction and fusion processing based on virtual maintenance action semantics on each virtual maintenance monitoring key frame in the time sequence of the virtual maintenance monitoring key frames respectively to obtain a time sequence of a virtual maintenance action semantic feature map;
Performing channel pruning optimization processing on each virtual maintenance action semantic feature map in the time sequence of the virtual maintenance action semantic feature map to obtain the time sequence of the optimized virtual maintenance action semantic feature map;
The optimized virtual maintenance action semantic feature images in the time sequence of the optimized virtual maintenance action semantic feature images are subjected to feature aggregation along the channel dimension and then input into a global feature perception fusion module, so that a virtual maintenance action semantic full-time domain associated feature image is obtained;
Based on the virtual maintenance action semantic full-time domain correlation feature map, determining whether the maintenance training operation action of the checked personnel has errors, determining whether to send out prompt information, and generating an check report.
Compared with the prior art, the virtual maintenance training assessment management method provided by the application has the advantages that the virtual maintenance monitoring video of the assessed personnel is collected through the camera, the virtual maintenance monitoring video is analyzed by introducing the data processing and image analysis algorithm based on artificial intelligence and machine vision at the rear end, so that the relation between the semantic meaning and time sequence association of the virtual maintenance action of the assessed personnel is captured, the irregular operation or the error action of the maintenance action is identified, and when the abnormal operation or the error action is detected, the prompt information is sent out, and the assessment report is generated. Therefore, more intelligent virtual maintenance training assessment management can be realized, and the maintenance action time sequence semantics of the assessment personnel are understood by performing overall analysis based on time sequence context on the action semantics of the assessment personnel, so that abnormal identification and error detection of maintenance training actions are performed, the application range and reliability of a virtual maintenance training assessment scheme are improved, and more objective and comprehensive assessment results are provided.
Drawings
The above and other objects, features and advantages of the present application will become more apparent by describing embodiments of the present application in more detail with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate the application and together with the embodiments of the application, and not constitute a limitation to the application. In the drawings, like reference numerals generally refer to like parts or steps.
FIG. 1 is a flow chart of a virtual maintenance training assessment management method according to an embodiment of the present application;
FIG. 2 is a data flow diagram of a virtual maintenance training assessment management method according to an embodiment of the present application.
Detailed Description
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
As used in the specification and in the claims, the terms "a," "an," "the," and/or "the" are not specific to a singular, but may include a plurality, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.
Although the present application makes various references to certain modules in a system according to embodiments of the present application, any number of different modules may be used and run on a user terminal and/or server. The modules are merely illustrative, and different aspects of the systems and methods may use different modules.
A flowchart is used in the present application to describe the operations performed by a system according to embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in order precisely. Rather, the various steps may be processed in reverse order or simultaneously, as desired. Also, other operations may be added to or removed from these processes.
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
In the conventional virtual maintenance training assessment management scheme, the assessment of the vehicle maintenance training assessment is generally performed by means of a standardized mode and a single rule, and the standardized assessment method is convenient for assessment and comparison, but can limit the creativity and flexibility of maintenance personnel. That is, in actual maintenance, the same fault may have multiple reasons and investigation methods, and conventional virtual training assessment often has difficulty in simulating such diversity, which may result in too single training content and operation mode. Therefore, in the actual maintenance training and checking of the vehicle, it is very important to flexibly apply various methods to troubleshoot the fault, which puts forward higher requirements and intellectualization for the virtual maintenance training and checking process. Accordingly, an optimized virtual repair training assessment management scheme is desired.
In the technical scheme of the application, a virtual maintenance training assessment management method is provided. FIG. 1 is a flow chart of a virtual maintenance training assessment management method according to an embodiment of the present application. FIG. 2 is a data flow diagram of a virtual maintenance training assessment management method according to an embodiment of the present application. As shown in fig. 1 and 2, the virtual maintenance training assessment management method according to the embodiment of the application includes the steps of:
S1, acquiring a virtual maintenance monitoring video of an inspected person;
S2, performing key frame sampling on the virtual maintenance monitoring video to obtain a time sequence of virtual maintenance monitoring key frames;
S3, performing multi-scale feature extraction and fusion processing based on the virtual maintenance action semantics on each virtual maintenance monitoring key frame in the time sequence of the virtual maintenance monitoring key frames to obtain the time sequence of the virtual maintenance action semantic feature map;
s4, performing channel pruning optimization processing on each virtual maintenance action semantic feature map in the time sequence of the virtual maintenance action semantic feature map to obtain the time sequence of the optimized virtual maintenance action semantic feature map;
S5, carrying out feature aggregation on each optimized virtual maintenance action semantic feature graph in the time sequence of the optimized virtual maintenance action semantic feature graph along the channel dimension, and then obtaining a virtual maintenance action semantic full-time domain associated feature graph through a global feature perception fusion module, and S6, determining whether a maintenance training operation action of an inspected person is wrong or not based on the virtual maintenance action semantic full-time domain associated feature graph, determining whether prompt information is sent or not, and generating an inspection report.
Particularly, in S1 and S2, virtual maintenance monitoring videos of the checked personnel are obtained, and key frame sampling is carried out on the virtual maintenance monitoring videos to obtain a time sequence of the virtual maintenance monitoring key frames. It should be understood that in the process of actually performing virtual maintenance training and assessment of a vehicle, the accuracy of operation and maintenance steps of an assessment person, the definition of analysis ideas of fault reasons, the standardization of use of tool equipment, the rationality of application of a fault screening method, the safety of equipment operation and the like are required to be monitored and assessed, and the difficulty is that the screening of fault points is non-unique, the screening methods used by different persons are different and cannot be uniformly standardized, and repair personnel are encouraged to flexibly use various methods to screen the fault points and repair equipment. Therefore, there is a need for efficient analysis of virtual repair surveillance videos to capture virtual repair action semantics and timing context correlation relationships of the inspected person. However, since the virtual maintenance monitoring video contains a large amount of redundant information and repeated semantics, the redundancy can affect the efficiency and the calculated amount of the subsequent virtual maintenance semantic analysis for the checked personnel. Based on this, in the technical solution of the present application, the key frame sampling needs to be performed on the virtual maintenance monitoring video to obtain the time sequence of the virtual maintenance monitoring key frame.
Specifically, in S3, multi-scale feature extraction and fusion processing based on the virtual repair action semantics is performed on each virtual repair monitoring key frame in the time sequence of virtual repair monitoring key frames to obtain the time sequence of virtual repair action semantic feature graphs. In the process of extracting the semantic features of the virtual maintenance time sequence and detecting the errors of the operation actions, because each virtual maintenance monitoring key frame contains the semantic information of the virtual maintenance actions of the personnel at a specific time point, the action semantics can be presented on different levels and different scales of the image frames, that is, in the maintenance actions, the features of different levels and different scales in the key frames can correspond to different operation details, such as gestures, tool use modes and the like. Thus, it is necessary to extract motion features at different levels and at different scales in the key frame. This is critical to understanding the complex actions and details. In the technical scheme of the application, each virtual maintenance monitoring key frame in the time sequence of the virtual maintenance monitoring key frame is respectively passed through a virtual maintenance action semantic feature extractor based on a pyramid information transmission network to obtain the time sequence of a virtual maintenance action semantic feature map.
It should be noted that, in the pyramid information delivery network, the bottom layer network is mainly responsible for extracting low-level features and space detail information of the image, while the high-level network is responsible for extracting more abstract high-level features. However, in the conventional network structure, the location information extracted by the lower layer network is difficult to directly influence the higher layer network, resulting in distortion of the location information in different levels of feature extraction. Furthermore, in pyramid information delivery networks, the delivery of features typically requires a significant amount of computation, especially when processing large-scale image key frame data. As the number of network layers increases, the amount of computation grows exponentially, which can lead to the training and reasoning process becoming very time consuming. Based on the above, in order to solve the problem that shallow position information in the pyramid information transmission network is difficult to influence deep features and causes a large amount of calculation amount required when feature information is transmitted, in the technical scheme of the application, each virtual maintenance monitoring key frame in the time sequence of virtual maintenance monitoring key frames is further respectively passed through a virtual maintenance action semantic feature extractor based on the pyramid information transmission network to obtain the time sequence of a virtual maintenance action semantic feature map. The virtual repair action semantic feature extractor is capable of extracting semantic information related to the repair action from the video, the features including not only the morphology of the action but also the intent of the action, which is critical to understanding the accuracy and rationality of the repair action. Particularly, in the technical scheme of the application, the virtual maintenance action semantic feature extractor based on the pyramid information transmission network comprises a feature extraction network and a feature fusion network, wherein in the feature fusion network, information transmission is realized by adopting a residual fusion and jump connection mode, so that shallow layer features and deep layer features in a virtual maintenance monitoring key frame are effectively fused, the position information extracted from the bottom layer can be effectively transmitted to a high layer, and the deep layer features can be better fused with the position information of the bottom layer in the key frame, thereby improving the expression capability and accuracy of the features. In addition, residual feature extraction and information transmission are carried out on virtual maintenance action features of different levels through a pyramid structure, so that the calculated amount in the feature transmission process is reduced, virtual maintenance semantic features of maintenance personnel are extracted on different scales, information is aggregated, efficient transmission and utilization of feature information are realized, meanwhile, unnecessary calculation cost is reduced, and the calculation efficiency and speed of a network are improved.
In the embodiment of the application, each virtual maintenance monitoring key frame in a time sequence of the virtual maintenance monitoring key frame is respectively processed by a virtual maintenance action semantic feature extractor based on a pyramid information transmission network to obtain a time sequence of a virtual maintenance action semantic feature map, the method comprises the steps of enabling the virtual maintenance monitoring key frame to be processed into a 3 x 3 convolution layer through a convolution kernel to obtain a virtual maintenance monitoring key frame shallow feature map, enabling the virtual maintenance monitoring key frame shallow feature map to be processed by a virtual maintenance action shallow residual information fusion enhancement module to obtain a virtual maintenance monitoring key frame shallow residual fusion feature map, enabling the virtual maintenance monitoring key frame shallow residual fusion feature map to be processed into a 3 x 3 convolution layer through the convolution kernel to obtain a virtual maintenance monitoring key frame deep feature map, enabling the virtual maintenance monitoring key frame deep feature map to be processed by a virtual maintenance action deep layer residual information fusion enhancement module to obtain a virtual maintenance monitoring key frame deep layer modulation feature map, enabling the virtual maintenance monitoring key frame deep feature map to be input into a SPPF layer based scale modulation module to obtain a virtual maintenance monitoring key frame deep modulation feature map, enabling the virtual maintenance monitoring key frame deep feature map to be processed by a virtual maintenance monitoring key frame deep layer deep feature map, and enabling the virtual maintenance monitoring key frame deep feature map to be processed by a virtual maintenance action deep feature map fusion enhancement module.
The virtual maintenance monitoring key frame shallow feature map is subjected to channel dimension reduction modulation through a point convolution layer to obtain a channel dimension reduction modulation virtual maintenance monitoring key frame shallow feature map, the channel dimension reduction modulation virtual maintenance monitoring key frame shallow feature map is subjected to feature extraction through a convolution layer with a convolution kernel of 3 multiplied by 3 to obtain a virtual maintenance monitoring key frame shallow hidden information feature map, the virtual maintenance monitoring key frame shallow hidden information feature map is subjected to channel dimension increase modulation through the point convolution layer to obtain a channel modulated virtual maintenance monitoring key frame shallow feature map, and the channel modulated virtual maintenance monitoring key frame shallow feature map and the channel dimension reduction modulation virtual maintenance monitoring key frame shallow feature map are fused to obtain the virtual maintenance monitoring key frame shallow hidden information fusion feature map.
The method comprises the steps of inputting a virtual maintenance monitoring key frame deep residual fusion characteristic map into a scale modulation module based on an SPPF layer to obtain a virtual maintenance monitoring key frame deep modulation characteristic map, obtaining a scale adjustment virtual maintenance monitoring key frame deep residual fusion characteristic map through the scale modulation module based on a point convolution layer, carrying out maximum value pooling treatment on each characteristic matrix along a channel dimension in the scale adjustment virtual maintenance monitoring key frame deep residual fusion characteristic map to obtain a virtual maintenance monitoring key frame deep residual fusion characteristic vector, and carrying out channel weighting reinforcement on the scale adjustment virtual maintenance monitoring key frame deep residual fusion characteristic map based on the virtual maintenance monitoring key frame deep residual fusion characteristic vector to obtain the virtual maintenance monitoring key frame deep modulation characteristic map.
Specifically, in S4, the channel pruning optimization is performed on each virtual repair action semantic feature map in the time sequence of virtual repair action semantic feature maps to obtain the time sequence of optimized virtual repair action semantic feature maps. Considering that each virtual maintenance action semantic feature map contains virtual maintenance action semantic multi-level fusion features about checked personnel at a specific time point, some of the features have significance for subsequent operation action abnormality and error detection tasks, some of the features are background interference or redundant information, and the contribution degree for subsequent detection tasks is low. Based on the above, in the technical scheme of the application, each virtual maintenance action semantic feature graph in the time sequence of the virtual maintenance action semantic feature graph is further subjected to channel pruning optimization processing to obtain the time sequence of the optimized virtual maintenance action semantic feature graph. Through channel pruning, channels with small contributions or redundancy to the model performance and the action error detection task of the subsequent maintenance training operation can be removed, so that the calculated amount is reduced, and the processing speed is increased. This is particularly important for real-time or near real-time virtual maintenance training assessment. In addition, the virtual maintenance action semantic feature map after pruning optimization is simpler, so that overfitting is avoided, and the generalization capability of the model on unknown data is improved. This means that the model can be more stably identified and evaluated in the face of a new repair scenario or a different repair action.
In the embodiment of the application, each virtual maintenance action semantic feature map in the time sequence of the virtual maintenance action semantic feature map is subjected to channel pruning optimization processing to obtain the time sequence of the optimized virtual maintenance action semantic feature map, the method comprises the steps of carrying out global averaging processing on each feature matrix along a channel dimension in the virtual maintenance action semantic feature map to obtain a virtual maintenance action semantic channel feature vector, multiplying the virtual maintenance action semantic channel feature vector with a weight matrix firstly and then carrying out position-wise addition on the weight matrix to obtain a virtual maintenance action semantic channel weight vector, masking the virtual maintenance action semantic channel weight vector to obtain a channel weight vector of the masking virtual maintenance action semantic feature map, carrying out weighted optimization on each feature matrix along the channel dimension in the masking virtual maintenance action semantic feature map to obtain the optimized virtual maintenance action semantic feature map, and carrying out mask processing on the virtual maintenance action semantic channel weight vector to obtain the channel weight vector of the masking virtual maintenance action semantic feature map, wherein the masking action semantic feature vector comprises taking each position feature value in the masking virtual maintenance action semantic feature map as a weighting coefficient, and taking the position feature value in the channel weight vector is equal to the preset feature value of the channel feature map, or the channel weight value is equal to the preset feature value in the channel weight value.
In summary, in the above embodiment, performing the channel pruning optimization processing on each virtual maintenance action semantic feature map in the time sequence of virtual maintenance action semantic feature maps to obtain the time sequence of optimized virtual maintenance action semantic feature maps includes performing the channel pruning optimization processing on each virtual maintenance action semantic feature map in the time sequence of virtual maintenance action semantic feature maps by using the following channel pruning optimization formula to obtain the time sequence of optimized virtual maintenance action semantic feature maps, where the formula is:
Wherein, For each virtual repair action semantic feature map,The global mean value pooling processing is carried out on each feature matrix along the channel dimension in the feature graph,For the virtual repair action semantic channel feature vector,As a matrix of weights, the weight matrix,As a result of the offset vector,Channel weight vectors for the virtual repair action semantic feature graphs,In order to mask the operation of the device,To mask the virtual repair action semantic channel weight vector,To perform a per-position point multiplication of the feature map along the channel dimension with each position feature value in the vector,And (5) a semantic feature map for the optimized virtual maintenance action.
Specifically, in S5, feature aggregation along the channel dimension is performed on each optimized virtual maintenance action semantic feature map in the time sequence of the optimized virtual maintenance action semantic feature map, and then the virtual maintenance action semantic full-time domain associated feature map is obtained through the global feature perception fusion module. It should be understood that, because each optimized virtual maintenance action semantic feature map includes virtual maintenance action semantic features after being optimized by saliency features at a specific time point, and there are time-sequence association relationships and interactions between the action semantics, in order to more fully and accurately understand the virtual maintenance action time sequence semantics and intentions of the examined personnel, the time-sequence context association relationships and interactions of the action semantics are utilized to perform subsequent maintenance training operation action error detection. The optimized virtual maintenance action semantic feature graphs at each time point are subjected to aggregation processing along the channel dimension, so that the maintenance action semantics and features can be integrated to enhance the time sequence relevance and the expression of the features. This helps to capture the overall trend and pattern of the repair action, not just local details, which is very useful for assessing the smoothness and logic of the repair action.
Specifically, after feature aggregation is performed on each optimized virtual maintenance action semantic feature graph along the channel dimension to form an optimized virtual maintenance action semantic aggregation feature graph in the whole time domain, global feature perception and time sequence correlation feature extraction are further performed on the feature graph, so that global time sequence correlation semantic feature information of virtual maintenance actions of checked personnel is captured, and understanding capacity of a model on context semantics of the whole time domain of the actions is improved. It is worth mentioning that the importance degree of the virtual maintenance action semantics at different time points can be automatically learned and captured through the processing of the global feature perception fusion module, namely the contribution degree of the subsequent maintenance training operation action detection and error recognition tasks is achieved, the virtual maintenance action semantics at different time sequences are adaptively weighted and enhanced, so that the whole virtual maintenance action semantics full-time domain associated feature expression is optimized, and a more accurate basis is provided for the subsequent maintenance training operation action error recognition.
In the embodiment of the application, each optimized virtual maintenance action semantic feature map in the time sequence of the optimized virtual maintenance action semantic feature map is subjected to feature aggregation along the channel dimension and then is subjected to a global feature perception fusion module to obtain a virtual maintenance action semantic full-time domain associated feature map, and the method comprises the steps of carrying out feature aggregation along the channel dimension on each optimized virtual maintenance action semantic feature map in the time sequence of the optimized virtual maintenance action semantic feature map to obtain an optimized virtual maintenance action semantic aggregation feature map, carrying out linear transformation on the optimized virtual maintenance action semantic aggregation feature map in the global feature perception fusion module to obtain a virtual maintenance action semantic linear transformation matrix, and passing the virtual maintenance action semantic linear transformation matrix through the global feature perception fusion modulePerforming soft maximum normalization processing in the function to obtain a virtual maintenance action semantic linear transformation weight matrix, performing position-wise multiplication on each feature matrix along the channel dimension in the optimized virtual maintenance action semantic aggregation feature map and the virtual maintenance action semantic linear transformation weight matrix in a global feature perception fusion module to obtain a virtual maintenance action semantic global channel feature map, and performing position-wise multiplication on the virtual maintenance action semantic global channel feature map in the global feature perception fusion moduleAnd multiplying the activated function by a trainable scaling factor after activation processing, and adding the activated function with the optimized virtual maintenance action semantic aggregation feature map according to positions to obtain a virtual maintenance action semantic full-time domain associated feature map.
The method comprises the steps of processing an optimized virtual maintenance action semantic aggregation feature map through a convolution layer to obtain a virtual maintenance action semantic feature matrix, multiplying the virtual maintenance action semantic feature matrix by a global perception weight matrix, and then adding the obtained weight modulation virtual maintenance action semantic feature matrix and the global perception bias matrix according to positions to obtain a virtual maintenance action semantic linear transformation matrix.
In summary, in the above embodiment, the processing of each optimized virtual maintenance action semantic feature map in the time sequence of the optimized virtual maintenance action semantic feature map through the global feature perception fusion module after feature aggregation along the channel dimension to obtain the virtual maintenance action semantic full-time domain associated feature map includes processing the optimized virtual maintenance action semantic aggregation feature map through the global feature perception fusion module according to the following global feature perception fusion formula to obtain the virtual maintenance action semantic full-time domain associated feature map, where the global feature perception fusion formula is:
Wherein, To aggregate feature graphs for post-optimization virtual repair action semantics,In order to perform a convolution operation on the feature map,For the matrix of global perceptual weights,For the global sense bias matrix,Representation ofThe function of the function is that,In order to multiply by location,A global channel feature map for virtual repair actions semantics,Representation ofThe function is activated and the function is activated,In order for the scaling factor to be trainable,In order to add by location,And (5) associating the characteristic diagram for the virtual maintenance action semantic full time domain.
Specifically, in S6, based on the semantic full-time domain correlation feature map of the virtual maintenance action, it is determined whether there is an error in the maintenance training operation action of the person under examination, and it is determined whether to issue a prompt message and generate an assessment report. In a specific example of the application, the virtual maintenance action semantic full-time domain associated feature map is passed through a classifier-based maintenance training operation action error identifier to obtain an identification result, and the identification result is used for indicating whether the maintenance training operation action of the checked person has an error. That is, the non-standard operation or the erroneous operation of the maintenance training operation is identified by performing the classification processing using the full-time domain related features of the virtual maintenance operation semantics. And responding to the recognition result that the error exists in the maintenance training operation action of the checked personnel, sending out prompt information and generating a check report. Therefore, more intelligent virtual maintenance training assessment management can be realized, the maintenance action time sequence semantics of the personnel can be understood by performing overall analysis based on time sequence context on the action semantics of the assessment personnel, so that abnormal identification and error detection of maintenance training actions are performed, the limitation of a single regular and standardized mode on the maintenance flexibility and diversity of the assessment personnel is avoided, the application range and reliability of a virtual maintenance training assessment scheme are improved, and more objective and comprehensive assessment results are provided.
Considering the time sequence of the virtual maintenance action semantic feature map to express the time sequence distribution of the image semantic features of the virtual maintenance monitoring key frame, when the channel pruning optimization processing and the global feature perception fusion after feature aggregation are carried out, the image semantic difference of each virtual maintenance action semantic feature map in the time sequence of the virtual maintenance action semantic feature map can cause the unbalance of attention weight distribution when the global feature perception fusion is carried out, so that the improvement of the detail image semantic aggregation expression effect of the virtual maintenance action semantic full-time domain associated feature map based on the fusion significant difference is expected.
Preferably, the step of operating the action error identifier to obtain the identification result through the maintenance training based on the classifier on the virtual maintenance action semantic full-time domain associated feature map comprises the following steps of:
Calculating the sum of absolute values of all feature values of the virtual maintenance action semantic full-time domain correlation feature map to obtain a first virtual maintenance action semantic full-time domain correlation and modulation value, and calculating the square root of the square sum of all feature values of the virtual maintenance action semantic full-time domain correlation feature map to obtain a second virtual maintenance action semantic full-time domain correlation and modulation value;
After carrying out point subtraction on the virtual maintenance action semantic full-time domain association feature map and the second virtual maintenance action semantic full-time domain association and modulation value, respectively carrying out point multiplication on the feature value number of the virtual maintenance action semantic full-time domain association feature map and the reciprocal of the first virtual maintenance action semantic full-time domain association and modulation value, and taking the reciprocal of each feature value to obtain a first virtual maintenance action semantic full-time domain association phase conversion feature map;
After carrying out point subtraction on the virtual maintenance action semantic full-time domain association feature map and the first virtual maintenance action semantic full-time domain association and modulation value, respectively carrying out point multiplication on the square root of the number of feature values of the virtual maintenance action semantic full-time domain association feature map and the reciprocal of the second virtual maintenance action semantic full-time domain association and modulation value, and taking the reciprocal of each feature value to obtain a second virtual maintenance action semantic full-time domain association phase conversion feature map;
subtracting the point multiplication feature map of the second virtual maintenance action semantic full-time domain associated phase conversion feature map from the point multiplication feature map of the weighted super parameter to obtain an optimized virtual maintenance action semantic full-time domain associated feature map;
And inputting the optimized virtual maintenance action semantic full-time domain associated feature map into a classifier-based maintenance training operation action error identifier to obtain an identification result.
Here, the virtual repair action semantic full time domain association feature mapThe optimization of (c) is expressed as:
Wherein, For a virtual repair action semantic full-time domain association feature map,Represents a set of real numbers,、、The width, the length and the height of the semantic full-time domain associated feature map of the virtual maintenance action are respectively represented,Representing the first time domain associated feature map of the virtual repair action semanticsThe characteristic value of the individual position is used,The number of feature values representing the semantic full-time domain associated feature map of the virtual repair action,Representing a first virtual maintenance action semantic full time domain association and modulation value,Representing a second virtual maintenance action semantic full time domain association and modulation value,Representing the multiplication by the position point,For the matrix multiplication to be performed,Indicating that the subtraction is performed by position,Representing the reciprocal of each feature value of the computed feature map,Representing a first virtual maintenance action semantic full-time domain associated phase conversion feature map,Representing a second virtual maintenance action semantic full-time domain associated phase conversion feature map,The weighted super-parameter is represented by a weighted super-parameter,And representing the optimized virtual maintenance action semantic full-time domain associated feature map.
In a preferred example, the difference of the feature value of the virtual maintenance action semantic full-time domain associated feature map relative to the difference of the whole feature set of the virtual maintenance action semantic full-time domain associated feature map and the modulation representation is used as semantic change intensity information, and the class phase conversion corresponding to the position-based intensity modulation is performed through the difference and the modulation representation form, so that the aggregation enhancement of semantic change phase perception can promote the axial aggregation receptive field along the feature aggregation direction through the space translation operation based on alternate stacking under the aggregation scale balance of the virtual maintenance action semantic full-time domain associated feature map, thereby promoting the perception effect of the aggregation semantic of the virtual maintenance action semantic full-time domain associated feature map on detail semantic change, promoting the expression effect of the virtual maintenance action semantic full-time domain associated feature map, and promoting the accuracy of the identification result obtained by the maintenance training operation error identifier based on the classifier. Therefore, irregular operation or error action of the virtual maintenance training operation action can be more accurately identified, prompt information is sent out and an assessment report is generated when abnormal operation or error action is detected, and therefore more intelligent virtual maintenance training assessment management is achieved.
In summary, the virtual maintenance training assessment management method according to the embodiment of the application is explained, which collects a virtual maintenance monitoring video of an assessed person through a camera, and introduces a data processing and image analysis algorithm based on artificial intelligence and machine vision into the rear end to analyze the virtual maintenance monitoring video, so as to capture the relation between the semantic meaning and time sequence association of the virtual maintenance action of the assessed person, thereby identifying the irregular operation or the error action of the maintenance action, and the like, and sending prompt information and generating an assessment report when detecting the abnormal operation or the error action. Therefore, more intelligent virtual maintenance training assessment management can be realized, and the maintenance action time sequence semantics of the assessment personnel are understood by performing overall analysis based on time sequence context on the action semantics of the assessment personnel, so that abnormal identification and error detection of maintenance training actions are performed, the application range and reliability of a virtual maintenance training assessment scheme are improved, and more objective and comprehensive assessment results are provided.
In one example, for the assessment of the operation class subjects, it is only necessary to perform condition detection on a certain operation step, that is, if the condition detection finds that the state is incorrect, the operation is judged to be incorrect, and the erroneous step is not scored. However, erroneous operations affecting the safety performance and the combat performance of the vehicle are directly recognized as bad. Meanwhile, in the aspect of scoring during the assessment, 1 score is added every 5% of the standard time in advance, 5 scores are obtained at most, 1 score is buckled every 1 minute overtime, and the system automatically requires to stop operation after 5 minutes overtime, so that the assessment score is directly determined as inequality. The assessment strategy is as follows:
(1) An operation step of searching the checking staff, if i=0, turning to "(6)";
(2) Each operation step is numbered according to the time sequence and is marked as ni, and the steps which can be synchronously operated are marked as mi;
(3) Judging whether each prescribed step has performed operation, whether the operation time is correct, whether related operation exists, if so, detecting whether the related operation is completed;
(4) If the operation time is incorrect (advanced or retarded), the deduction is not carried out according to the operation time;
(5) If the related operation of the operation is not completed, the operation cannot be continued, and meanwhile, the score of the step is deducted, if the operation is completed, the operation can be continued, and whether the operation is correctly scored or not is judged, if the related operation belongs to a sequential error, the deduction is carried out according to the sequential error, and if the related operation belongs to a disordered type, the deduction is carried out according to a missed operation error;
(6) And (5) calculating the assessment score by integrating the operation safety and the operation duration, and ending.
The repair subjects training is implemented by links such as material preparation, fault finding, fault cause analysis, fault troubleshooting, repair, verification, vehicle recovery and the like, and different scoring modes are adopted due to different assessment key points of each link.
1. Scoring method for material preparation links
And a material preparation link, which mainly checks whether personnel prepare the tool equipment sufficiently, whether the tool equipment is properly selected and whether the using method is proper. Thus, scoring should be done in a manner that scores the importance of the tool equipment.
2. Scoring method for fault finding link
And finding out a fault link, and mainly checking whether personnel equipment is correctly operated or not and judging whether a fault phenomenon can be accurately judged. Therefore, the operation assessment of a certain type of vehicle can be scored by adopting the method, and if no fault phenomenon is found, the vehicle is directly determined as failed.
3. Fault cause analysis link scoring method
The analysis of the fault cause is the precondition of the implementation of the follow-up fault diagnosis and repair links, directly determines the effect of fault elimination, and mainly checks whether the thought of the fault analysis of repair personnel is clear and the sequence of fault point investigation is reasonable, so that higher score weight is applied. Meanwhile, the determination of the fault points and the fault checking sequence is different from person to person, and no sequence exists, so that the fault point sequences selected by the checking personnel are scored. For example, the display displays the failure phenomenon of "XXX failure", and possible failure points thereof include cables, push-button switches, and electronic boxes. Wherein the fault probability of the electronic box is the largest, the next is the button switch, the optimal order of the electronic box, the button switch and the cable is the highest score of the order of the selected order of the electronic box, the button switch and the cable. The specific scoring algorithm is as follows:
Assuming that n possible fault points of a certain fault phenomenon are respectively marked as F1,..and Fn, the occurrence probability of each fault is determined empirically and respectively marked as p1, p2,..pn, the ordering positions of the fault points are marked as S1 from front to back..sn, and the probability difference Dij of the Sj fault points arranged by the current bit Si and the subsequent bit is calculated sequentially from the S1 position.
Dij=1 pSi>=pSj
Dij=0 pSi<pSj
The score Rmax represents the maximum value of R, the full score of the partial score is Cmax, the final score of the ranking c=r×cmax/Rmax, the final score 4 score 5.
4. Grading method for fault investigation link
The fault checking is mainly carried out item by item according to fault points determined by fault analysis, and the capability of checking and positioning by repairing personnel by using tools, instruments and detection equipment is examined. Because the troubleshooting of the fault points has no time sequence requirement, and repair staff is encouraged to quickly locate the fault points by using the least detection means, the troubleshooting of the fault points is not sequenced, the troubleshooting method steps of each fault point are mainly scored, and the difficulty, the complexity and the fault probability of each fault point troubleshooting are considered when the assigning is performed, so that different score weights are given. Taking the example of a shooter display showing a fault of "XXX fault", possible fault points include cables, push-button switches, electronic boxes. In the fault subject examination, the fault investigation link is endowed with a score of 25, wherein the fault probability of the electronic box is the largest, and then the electronic box is provided with a button switch and a cable, so that the fault investigation score of the electronic box is 15 minutes, wherein the electronic box is detected by a detection device for 10 minutes, the fault is detected and judged to be 5 minutes to a circuit board by a detector, and the fault investigation scores of the cable and the button switch are all 5 minutes. When each fault point is examined, scoring is carried out according to operation steps, and the adopted method is the same as that of the operation type examination subject.
5. Repair link scoring method
Repair, mainly adopt the repair mode of changing a part to repair the trouble spot that is found through trouble shooting, examine repair personnel and utilize the maintenance tool to carry out partial subassembly decomposition, spare part change, assembly and debugging ability. The repair is to disassemble and assemble the parts according to the sequence required in the regulations, the whole repair flow has time sequence requirement and part of operation has no sequence requirement, therefore, the condition detection is adopted for a certain operation step, if the condition detection finds that the state is incorrect, the operation is judged to be wrong, and the wrong step is not scored. Taking a shooter display as an example for displaying XXX fault, wherein the fault point is an AP02 board of the electronic box, and the repairing step comprises the steps of firstly, detaching a top cover of the electronic box by using a cross screw driver, then loosening an AP02 locking mechanism by using a straight screw driver, pulling out the AP02, and replacing a spare part AP02 circuit board. The operation steps are executed strictly in sequence, when a certain step operation is performed, whether the previous step is completed is detected, if the previous step is completed, the score is obtained, and if the previous step is not completed, the operation can be continued, and the previous step operation is not given.
6. Team collaborative operation assessment and evaluation method
The team cooperation assessment mainly aims at the content that a plurality of teams need to conduct fault investigation, repair and the like in the repair subjects, and besides the individual operation and repair capability, the cooperation mercy degree among the teams needs to be assessed, and the team cooperation assessment is mainly used for scoring through time consumption. In the repair subjects, for possible fault points, a plurality of fault points can be synchronously subjected to fault investigation and repair, or a plurality of fault points can be simultaneously subjected to fault investigation and repair, so that a team leader in a team is required to carry out task allocation to determine which cooperative mode is adopted, and the investigation and repair of each fault point are independent of each other and have no time sequence, therefore, the adopted grading method still depends on the steps of the investigation and repair methods of each fault point, the grading is carried out according to the operation steps, the software carries out condition detection on one operation step, the operation error is judged when the condition detection finds that the state is incorrect, and the wrong step is not scored without considering the identity of the repair personnel. The assessment and evaluation software only needs to solve the problem that multiple persons simultaneously operate the vehicle, the solution method of the project is to adopt a multithread parallel operation mode for two unassociated fault points, score the two unassociated fault points independently, and the fault points with association relation cannot be simultaneously checked and repaired by the multiple persons. Taking a shooter display to display XXX fault as an example, possible fault points comprise a cable, a button switch and an electronic box, wherein the cable is connected with the electronic box, the cable and the electronic box have an association relation, the cable must be detached firstly to conduct fault investigation and repair on the electronic box and the cable respectively, and if a repair person does not detach the cable at this moment, namely, the fire box is detached, software considers that the operation step is wrong in operation time and the operation step should be deducted.
The foregoing description of the embodiments of the present disclosure has been presented for purposes of illustration and description, and is not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the various embodiments described. The terminology used herein was chosen in order to best explain the principles of the embodiments, the practical application, or the improvement of technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (9)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411390168.8A CN119445650A (en) | 2024-10-08 | 2024-10-08 | Virtual maintenance training assessment management method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411390168.8A CN119445650A (en) | 2024-10-08 | 2024-10-08 | Virtual maintenance training assessment management method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN119445650A true CN119445650A (en) | 2025-02-14 |
Family
ID=94521048
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411390168.8A Pending CN119445650A (en) | 2024-10-08 | 2024-10-08 | Virtual maintenance training assessment management method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119445650A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140219666A1 (en) * | 2011-03-03 | 2014-08-07 | Technion Research And Development Foundation Ltd. | Coherent and self-coherent signal processing techniques |
| CN117574596A (en) * | 2023-08-30 | 2024-02-20 | 北京海鹰科技情报研究所 | Equipment maintenance demonstration simulation method based on virtual reality |
| CN117975341A (en) * | 2024-03-25 | 2024-05-03 | 集美大学 | Fishing vessel operation behavior recognition method based on deep learning |
| CN118537941A (en) * | 2024-07-29 | 2024-08-23 | 长春职业技术学院 | Attendance intelligent recognition system and method based on target detection and face recognition |
| CN118629635A (en) * | 2024-08-13 | 2024-09-10 | 吉林大学 | System and method for predicting the therapeutic effect of deep vein thrombosis thrombolysis |
| CN118674715A (en) * | 2024-08-22 | 2024-09-20 | 吉林大学 | Digestive endoscopy quality control system and method based on artificial intelligence |
-
2024
- 2024-10-08 CN CN202411390168.8A patent/CN119445650A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140219666A1 (en) * | 2011-03-03 | 2014-08-07 | Technion Research And Development Foundation Ltd. | Coherent and self-coherent signal processing techniques |
| CN117574596A (en) * | 2023-08-30 | 2024-02-20 | 北京海鹰科技情报研究所 | Equipment maintenance demonstration simulation method based on virtual reality |
| CN117975341A (en) * | 2024-03-25 | 2024-05-03 | 集美大学 | Fishing vessel operation behavior recognition method based on deep learning |
| CN118537941A (en) * | 2024-07-29 | 2024-08-23 | 长春职业技术学院 | Attendance intelligent recognition system and method based on target detection and face recognition |
| CN118629635A (en) * | 2024-08-13 | 2024-09-10 | 吉林大学 | System and method for predicting the therapeutic effect of deep vein thrombosis thrombolysis |
| CN118674715A (en) * | 2024-08-22 | 2024-09-20 | 吉林大学 | Digestive endoscopy quality control system and method based on artificial intelligence |
Non-Patent Citations (2)
| Title |
|---|
| JUJI OU等: "Research and Explainable Analysis of a Real-Time Passion Fruit Detection Model Based on FSOne-YOLOv7", 《AGRONOMY》, 23 July 2023 (2023-07-23), pages 1 - 28 * |
| 黎国斌等: "基于多尺度特征融合的YOLOv3行人检测算法", 《 计算机与数字工程》, vol. 52, no. 01, 20 January 2024 (2024-01-20), pages 145 - 149 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106201871B (en) | Based on the Software Defects Predict Methods that cost-sensitive is semi-supervised | |
| US8868985B2 (en) | Supervised fault learning using rule-generated samples for machine condition monitoring | |
| CN111695620B (en) | Method and system for detecting and correcting time sequence abnormal data of power system | |
| CN116934304B (en) | Intelligent power distribution room equipment operation maintenance management system and method thereof | |
| CN117520924B (en) | Island photovoltaic operation and maintenance fault cause analysis method and system based on multimodal data | |
| CN115309575B (en) | Microservice fault diagnosis method, device and equipment based on graph convolutional neural network | |
| CN111401418A (en) | Employee dressing specification detection method based on improved Faster r-cnn | |
| CN108647707B (en) | Probabilistic neural network creation method, failure diagnosis method and apparatus, and storage medium | |
| CN113676343B (en) | Fault source positioning method and device for power communication network | |
| CN112906764A (en) | Communication safety equipment intelligent diagnosis method and system based on improved BP neural network | |
| Huang et al. | Attack detection and data generation for wireless cyber-physical systems based on self-training powered generative adversarial networks | |
| KR102760440B1 (en) | Artificial intelligence-based radiographic inspection system | |
| CN119445650A (en) | Virtual maintenance training assessment management method | |
| CN117520942A (en) | Fault diagnosis method and device based on bilinear convolutional neural network | |
| CN116071624B (en) | Smoking detection data labeling method based on active learning | |
| CN116127324A (en) | Fault detection method and device for key components in uninterruptible power supply | |
| CN114187292B (en) | Abnormality detection method, apparatus, device and storage medium for cotton spinning paper tube | |
| de Araujo et al. | Explainability for Property Violations in Cyberphysical Systems: An Immune-Inspired Approach | |
| CN113537407B (en) | Image data evaluation processing method and device based on machine learning | |
| CN114510431B (en) | Workload-aware intelligent contract defect prediction method, system and equipment | |
| CN118167426B (en) | Intelligent monitoring equipment and method for mine safety management | |
| CN113419140B (en) | Distribution network line fault diagnosis method and device considering human-computer-object cooperation | |
| CN118537808B (en) | Power equipment monitoring method and system based on image processing | |
| CN119415221B (en) | Active test method, device, equipment, medium and product of AI computing force cluster | |
| CN111160454B (en) | Quick change signal detection method and device |
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 |