Digital asset management method based on environmental protection field and application

Through machine vision technology based on deep learning, semantic analysis and context-related enhancement of environmental monitoring videos, the problem of difficulty in extracting key information in traditional environmental data management is solved, and automatic archiving and efficient management of digital assets in the field of environmental protection is realized.

CN118968420AInactive Publication Date: 2024-11-15ZHENGZHOU FUMING ENVIRONMENT PROTECTION & TECH CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411184545.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional environmental data management methods are difficult to effectively extract and utilize key information in surveillance videos, and cannot meet the needs of large-scale and real-time environmental data management and analysis.

Method used

Using machine vision technology based on deep learning to perform semantic analysis of environmental monitoring videos, extract image semantic features of keyframes, and mine the full-time domain environmental status information through context semantic correlation enhancement module to realize automatic archiving of digital assets in the field of environmental protection and manage.

Benefits of technology

It improves the management efficiency of environmental data, reduces the demand for manual operations, and improves the accuracy and real-timeness of data management.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118968420A_ABST
    Figure CN118968420A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of data management, and particularly discloses a digital asset management method based on the field of environmental protection and application, and the method comprises the steps: carrying out the semantic analysis of an environment monitoring video imported by a user through employing a machine vision technology based on deep learning, extracting the image semantic features of each key frame in the environment monitoring video, and carrying out the recognition of the image semantic features; and context semantic association enhancement is carried out on the digital assets to mine full-time-domain environment state information, so that the type of the environment state is intelligently identified, and automatic filing and management of the digital assets in the field of environmental protection are realized. Therefore, the management efficiency of the environmental data can be improved, the requirement of manual operation is reduced, and the accuracy and the real-time performance of data management are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data management technology, and more specifically, to a digital asset management method and application based on the field of environmental protection. Background Art

[0002] With the increasing awareness of environmental protection in society, the monitoring and management of the ecological environment has become an indispensable part of today's society. In order to effectively promote environmental protection work and support the formulation of various decisions and scientific research, a large amount of environmental data and information must be managed rigorously and meticulously, including but not limited to the collation, storage and analysis of environmental quality monitoring data, pollution source emission data, ecosystem change data and other types of information related to the natural environment.

[0003] Traditional environmental data management methods often rely on manual operations, with staff collecting, collating, analyzing, and archiving environmental data. In the field of environmental protection, surveillance video is an important data source for obtaining environmental status information. However, due to the characteristics of surveillance video, such as large data volume, complex information, a wide variety, and high real-time nature, traditional manual management methods are difficult to effectively extract and utilize key information, and cannot meet the needs of large-scale, real-time environmental data management and analysis. Therefore, an optimized digital asset management method based on the field of environmental protection is expected. Summary of the invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a digital asset management method and application based on the field of environmental protection, which uses machine vision technology based on deep learning to perform semantic analysis on environmental monitoring videos imported by users, extracts the image semantic features of each key frame in the environmental monitoring video, and strengthens its contextual semantic association to mine the full-time domain environmental status information, thereby intelligently identifying the type of environmental status and realizing automatic archiving and management of digital assets in the field of environmental protection. In this way, the management efficiency of environmental data can be improved, the need for manual operation can be reduced, and the accuracy and real-time performance of data management can be improved.

[0005] Accordingly, according to one aspect of the present application, a digital asset management method based on the field of environmental protection is provided, which includes:

[0006] Acquire a digital asset imported by a user, wherein the digital asset is an environmental monitoring video;

[0007] Extracting environmental state features of the environmental monitoring video to obtain a sequence of environmental state semantic feature graphs;

[0008] Inputting each of the environmental state semantic feature maps in the sequence of environmental state semantic feature maps into a feature excitation-squeezing module including a feature multi-scale perception enhancement layer and a multi-scale semantic feature extraction layer to obtain a sequence of environmental state semantic enhancement feature vectors;

[0009] Inputting the sequence of the environment state semantic enhancement feature vectors into a contextual semantic association enhancement module based on saliency-globality to obtain a sequence of association enhanced environment state semantic feature vectors;

[0010] Based on the sequence of the associated enhanced environment state semantic feature vectors, the digital assets imported by the user are automatically archived.

[0011] In the above-mentioned digital asset management method based on the field of environmental protection, the environmental state characteristics of the environmental monitoring video are extracted to obtain a sequence of environmental state semantic feature maps, including: key frame sampling of the environmental monitoring video to obtain a sequence of environmental monitoring key frames; each environmental monitoring key frame in the sequence of environmental monitoring key frames is passed through an environmental state feature extractor based on a void convolutional neural network model to obtain a sequence of environmental state semantic feature maps.

[0012] In the above-mentioned digital asset management method based on the field of environmental protection, each environmental state semantic feature map in the sequence of environmental state semantic feature maps is respectively input into a feature excitation-squeezing module including a feature multi-scale perception enhancement layer and a multi-scale semantic feature extraction layer to obtain a sequence of environmental state semantic enhancement feature vectors, including: each environmental state semantic feature map in the sequence of environmental state semantic feature maps is subjected to feature excitation through the feature multi-scale perception enhancement layer of the feature excitation-squeezing module to obtain a sequence of environmental state perception enhancement semantic feature maps; each environmental state perception enhancement semantic feature map in the sequence of environmental state perception enhancement semantic feature maps is subjected to feature compression through the multi-scale semantic feature extraction layer of the feature excitation-squeezing module to obtain a sequence of environmental state semantic enhancement feature vectors.

[0013] In the above-mentioned digital asset management method based on the field of environmental protection, each environmental state semantic feature map in the sequence of environmental state semantic feature maps is subjected to feature excitation through the feature multi-scale perception enhancement layer of the feature excitation-squeezing module to obtain a sequence of environmental state perception enhancement semantic feature maps, including: inputting the environmental state semantic feature map into the low-frequency detection branch of the feature multi-scale perception enhancement layer to obtain a local low-frequency activation semantic feature vector of the environmental state, wherein the low-frequency detection branch includes a first point convolution layer, a first global mean pooling layer and a first nonlinear activation layer; inputting the environmental state semantic feature map into the precision of the feature multi-scale perception enhancement layer; A simplified compression branch is formed to obtain a channel compressed semantic feature map of the environment state, wherein the simplified compression branch includes a second point convolution layer; the environment state semantic feature map is input into the global space expansion perception branch of the feature multi-scale perception enhancement layer to obtain an environment state receptive field expansion global activation semantic feature matrix, wherein the global space expansion perception branch includes a hole convolution layer, a third point convolution layer and a second nonlinear activation layer; the environment state local low-frequency activation semantic feature vector, the environment state channel compressed semantic feature map and the environment state receptive field expansion global activation semantic feature matrix are feature superimposed and integrated to obtain the environment state perception enhancement semantic feature map.

[0014] In the above-mentioned digital asset management method based on the field of environmental protection, the local low-frequency activation semantic feature vector of the environmental state, the channel compression semantic feature map of the environmental state and the global activation semantic feature matrix of the receptive field expansion of the environmental state are feature superimposed and integrated to obtain the environmental state perception enhancement semantic feature map, including: performing positional point multiplication on the environmental state receptive field expansion global activation semantic feature matrix and each feature matrix along the channel dimension of the channel compression semantic feature map of the environmental state to obtain a channel-modulated environmental state global semantic feature map; using each eigenvalue in the local low-frequency activation semantic feature vector of the environmental state as a weight, performing point multiplication on each feature matrix along the channel dimension of the channel compression semantic feature map of the environmental state to obtain a channel-modulated environmental state local low-frequency semantic feature map; adding the channel-modulated environmental state global semantic feature map and the channel-modulated environmental state local low-frequency semantic feature map by position to obtain an environmental state multi-scale semantic fusion feature map; performing hole convolution coding on the environmental state multi-scale semantic fusion feature map to obtain the environmental state perception enhancement semantic feature map.

[0015] In the above-mentioned digital asset management method based on the field of environmental protection, each environmental state perception enhanced semantic feature map in the sequence of environmental state perception enhanced semantic feature maps is subjected to feature compression through the multi-scale semantic feature extraction layer of the feature excitation-compression module to obtain a sequence of environmental state semantic enhanced feature vectors, including: upsampling the environmental state perception enhanced semantic feature map to obtain an upsampled environmental state perception enhanced semantic feature map; passing the upsampled environmental state perception enhanced semantic feature map through the global semantic feature extraction branch of the multi-scale semantic feature extraction layer to obtain an environmental state perception enhanced global semantic feature vector. The global semantic feature extraction module comprises a second global mean pooling layer, a fourth point convolution layer, a first batch normalization layer and a third non-linear activation layer; the upsampled environment state perception enhanced semantic feature map is passed through the local semantic feature extraction branch of the multi-scale semantic feature extraction layer to obtain an environment state perception enhanced local semantic feature vector, wherein the local semantic feature extraction module comprises a fifth point convolution layer, a second batch normalization layer and a fourth non-linear activation layer; the environment state perception enhanced global semantic feature vector and the environment state perception enhanced local semantic feature vector are fused to obtain the environment state semantic enhanced feature vector.

[0016] In the above-mentioned digital asset management method based on the field of environmental protection, the sequence of the environmental state semantic enhancement feature vectors is input into a contextual semantic association enhancement module based on significance-globality to obtain a sequence of association-enhanced environmental state semantic feature vectors, including: calculating the maximum value of each environmental state semantic enhancement feature vector in the sequence of the environmental state semantic enhancement feature vectors to obtain an environmental state semantic salient feature vector composed of each of the maximum values; calculating the average value of each environmental state semantic enhancement feature vector in the sequence of the environmental state semantic enhancement feature vectors to obtain an environmental state semantic global feature vector composed of each of the average values; performing one-dimensional convolution encoding, nonlinear activation and linear transformation on the environmental state semantic salient feature vector and the environmental state semantic global feature vector to obtain an environmental state semantic salient feature contextual association activation feature vector and an environmental state semantic global feature contextual association activation feature vector; based on the environmental state semantic salient feature contextual association activation feature vector and the environmental state semantic global feature contextual association activation feature vector, performing semantic association enhancement on the sequence of the environmental state semantic enhancement feature vectors to obtain the sequence of the association-enhanced environmental state semantic feature vectors.

[0017] In the above-mentioned digital asset management method based on the field of environmental protection, based on the environmental state semantic prominent feature context associated activation feature vector and the environmental state semantic global feature context associated activation feature vector, the sequence of the environmental state semantic enhanced feature vectors is semantically enhanced to obtain the sequence of the associated enhanced environmental state semantic feature vectors, including: calculating the positional addition of the environmental state semantic prominent feature context associated activation feature vector and the environmental state semantic global feature context associated activation feature vector to obtain the environmental state prominent-global feature context associated activation feature vector; performing nonlinear activation based on the tanh activation function and nonlinear activation based on the sigmoid activation function on the environmental state prominent-global feature context associated activation feature vector to obtain a first activation result and a second activation result, and performing positional point multiplication on the first activation result and the second activation result to obtain an environmental state prominent-global feature context associated weighted feature vector; using the feature values ​​of each position in the environmental state prominent-global feature context associated weighted feature vector as the weight, performing vector-by-vector weighting on the sequence of the environmental state semantic enhanced feature vectors, and then adding and fusing them with the sequence of the environmental state semantic enhanced feature vectors to obtain the sequence of the associated enhanced environmental state semantic feature vectors.

[0018] In the above-mentioned digital asset management method based on the field of environmental protection, the digital assets imported by the user are automatically archived based on the sequence of the associated enhanced environmental state semantic feature vectors, including: after cascading the sequence of the associated enhanced environmental state semantic feature vectors into a full-time domain environmental state semantic association feature vector, the full-time domain environmental state semantic association feature vector is input into a classifier-based digital asset classifier to obtain a classification result, and the classification result is used to represent the type label to which the digital asset belongs; based on the classification result, the digital assets imported by the user are automatically archived.

[0019] According to another aspect of the present application, an application of the above-mentioned digital asset management method in environmental protection is provided.

[0020] Compared with the prior art, the digital asset management method and application based on the field of environmental protection provided by this application uses machine vision technology based on deep learning to perform semantic analysis on the environmental monitoring video imported by the user, extracts the image semantic features of each key frame in the environmental monitoring video, and strengthens its contextual semantic association to mine the full-time domain environmental status information, thereby intelligently identifying the type of environmental status and realizing automatic archiving and management of digital assets in the field of environmental protection. In this way, the management efficiency of environmental data can be improved, the need for manual operation can be reduced, and the accuracy and real-time performance of data management can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 It is a flowchart of a digital asset management method based on the environmental protection field according to an embodiment of the present application.

[0023] Figure 2 It is a schematic diagram of the architecture of a digital asset management method based on the environmental protection field according to an embodiment of the present application.

[0024] Figure 3 It is a flowchart of a sequence of extracting environmental status features of the environmental monitoring video to obtain environmental status semantic feature graphs in a digital asset management method based on the environmental protection field according to an embodiment of the present application.

[0025] Figure 4 The present invention is a flowchart of inputting each environmental state semantic feature map in the sequence of environmental state semantic feature maps into a feature excitation-squeezing module including a feature multi-scale perception enhancement layer and a multi-scale semantic feature extraction layer in the digital asset management method based on the environmental protection field according to an embodiment of the present application to obtain a sequence of environmental state semantic enhancement feature vectors.

[0026] Figure 5 This is a flowchart of a method for digital asset management based on the field of environmental protection according to an embodiment of the present application, in which each environmental state semantic feature graph in the sequence of environmental state semantic feature graphs is subjected to feature excitation through the feature multi-scale perception enhancement layer of the feature excitation-squeezing module to obtain a sequence of environmental state perception enhanced semantic feature graphs.

[0027] Figure 6 This is a flowchart of the method for digital asset management based on the field of environmental protection according to an embodiment of the present application, in which each environmental state perception enhanced semantic feature map in the sequence of environmental state perception enhanced semantic feature maps is subjected to feature compression through the multi-scale semantic feature extraction layer of the feature excitation-compression module to obtain a sequence of environmental state semantic enhanced feature vectors.

[0028] Figure 7 The present invention is a flowchart of inputting the sequence of environmental state semantic enhancement feature vectors into a contextual semantic association enhancement module based on significance-globality in a digital asset management method based on the field of environmental protection according to an embodiment of the present application to obtain a sequence of association-enhanced environmental state semantic feature vectors.

[0029] Figure 8 The present invention is a flowchart of automatically archiving the digital assets imported by the user based on the sequence of the associated enhanced environmental state semantic feature vectors in the digital asset management method based on the environmental protection field according to an embodiment of the present application. DETAILED DESCRIPTION

[0030] Below, the embodiments of the present application will be described in more detail in conjunction with the accompanying drawings, and the above and other purposes, features and advantages of the present application will become more apparent. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein.

[0031] As mentioned in the above background technology, due to the characteristics of large data volume, complex information, wide variety, and high real-time nature of surveillance videos, traditional manual management methods are difficult to effectively extract and utilize key information therein, and cannot meet the needs of large-scale, real-time environmental data management and analysis. In response to the above technical problems, the technical concept of this application is to use machine vision technology based on deep learning to perform semantic analysis on environmental surveillance videos imported by users, extract the image semantic features of each key frame in the environmental surveillance video, and enhance its contextual semantic association to mine full-time domain environmental status information, thereby intelligently identifying the type of environmental status and realizing automatic archiving and management of digital assets in the field of environmental protection. In this way, the management efficiency of environmental data can be improved, the need for manual operation can be reduced, and the accuracy and real-time nature of data management can be improved.

[0032] Figure 1 It is a flowchart of a digital asset management method based on the environmental protection field according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of a digital asset management method based on the environmental protection field according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the digital asset management method based on the field of environmental protection includes the following steps: S110, obtaining a digital asset imported by a user, wherein the digital asset is an environmental monitoring video; S120, extracting the environmental state features of the environmental monitoring video to obtain a sequence of environmental state semantic feature maps; S130, inputting each environmental state semantic feature map in the sequence of environmental state semantic feature maps into a feature excitation-squeezing module including a feature multi-scale perception enhancement layer and a multi-scale semantic feature extraction layer to obtain a sequence of environmental state semantic enhancement feature vectors; S140, inputting the sequence of environmental state semantic enhancement feature vectors into a contextual semantic association enhancement module based on saliency-globality to obtain a sequence of association-enhanced environmental state semantic feature vectors; S150, automatically archiving the digital assets imported by the user based on the sequence of association-enhanced environmental state semantic feature vectors.

[0033] In the above-mentioned digital asset management method based on the field of environmental protection, the step S110 obtains the digital assets imported by the user, wherein the digital assets are environmental monitoring videos. It should be understood that environmental monitoring videos usually contain a large amount of spatiotemporal information, such as lighting, weather, vegetation coverage, biological activities, etc., which directly reflect the state and changes of the actual environment and are the most intuitive and basic data source in environmental protection work. In the technical solution of the present application, by obtaining the digital asset of environmental monitoring video and analyzing and processing it using deep learning technology, the dynamic change information of the environmental state can be captured, thereby realizing its intelligent archiving management.

[0034] In the above-mentioned digital asset management method based on the field of environmental protection, the step S120 extracts the environmental state features of the environmental monitoring video to obtain a sequence of environmental state semantic feature graphs. Figure 3 The flowchart is a sequence of extracting the environmental state features of the environmental monitoring video to obtain the environmental state semantic feature graph based on the digital asset management method in the field of environmental protection according to the embodiment of the present application. Figure 3 As shown, the step S120 includes: S121, performing key frame sampling on the environment monitoring video to obtain a sequence of environment monitoring key frames; S122, respectively passing each environment monitoring key frame in the sequence of environment monitoring key frames through an environment state feature extractor based on a void convolutional neural network model to obtain a sequence of environment state semantic feature graphs.

[0035] Specifically, the step S121 performs key frame sampling on the environment monitoring video to obtain a sequence of environment monitoring key frames. It should be understood that since the environment monitoring video usually has a large number of frames and a long playback time, and contains a large amount of redundant information, directly processing the entire video data is not only inefficient, but may also cause unnecessary interference to the extraction of environmental state change characteristics. Therefore, in order to improve the efficiency and accuracy of video processing, in the technical solution of the present application, a key frame sampling method is adopted to sample the environment monitoring video at a predetermined sampling frequency, and key frames are extracted therefrom to reduce the interference of redundant information. Compared with processing the entire video stream, the key frame sampling method can greatly reduce the computational complexity, while retaining sufficient environmental state information for subsequent analysis, significantly reducing the consumption of computing resources, so that the environment monitoring video can be analyzed more efficiently and the changes in environmental state can be understood more accurately under limited computing resources.

[0036] Specifically, in step S122, each environment monitoring key frame in the sequence of the environment monitoring key frames is respectively subjected to an environment state feature extractor based on a dilated convolutional neural network model to obtain a sequence of the environment state semantic feature graph. That is, in order to fully extract the image information of the environment monitoring key frames to form a comprehensive understanding of the environment state, in the technical solution of the present application, an environment state feature extractor based on a dilated convolutional neural network model is used to perform image processing on each environment monitoring key frame. It should be known to those skilled in the art that the dilated convolutional neural network model (DCNN) increases the receptive field by introducing holes (i.e., zero values) in the convolution kernel without increasing the number of parameters and computational complexity, so that the network can capture a wider range of spatial information. Therefore, the environment state feature extractor, based on the dilated convolution characteristics, can effectively capture a wide range of spatial context information in each environment monitoring key frame, such as biological status, vegetation coverage, etc., and convert the pixel information in the image into a higher level of semantic feature representation, providing a strong data basis for subsequent environment state type recognition processing.

[0037] In the above-mentioned digital asset management method based on the field of environmental protection, in step S130, each environmental state semantic feature map in the sequence of environmental state semantic feature maps is respectively input into a feature excitation-squeezing module including a feature multi-scale perception enhancement layer and a multi-scale semantic feature extraction layer to obtain a sequence of environmental state semantic enhancement feature vectors. It should be understood that in the field of environmental protection, the multi-scale characteristics and complex semantic information of environmental monitoring videos are important bases for analyzing changes in environmental states. However, feature extraction of a single scale is often difficult to fully capture the diversity and complexity of environmental states. Therefore, in the technical solution of the present application, a feature excitation-squeezing module including a feature multi-scale perception enhancement layer and a multi-scale semantic feature extraction layer is introduced to process each environmental state semantic feature map separately. Specifically, the feature excitation-squeezing module first performs multi-scale feature perception and fusion on the environmental state semantic feature map through the feature multi-scale perception enhancement layer, and further expands the perception range of the multi-scale fusion feature through the dilated convolution, thereby achieving comprehensive enhancement and accurate expression of environmental state features. Then, the feature representation output by the feature multi-scale perception enhancement layer is sent to the multi-scale semantic feature extraction layer, and through further convolution and pooling operations, the local detail semantic information and global context semantic information of the environment state semantic feature map are extracted and fused, and the feature dimension is compressed to obtain a more comprehensive and accurate description of the environment state feature. In this way, the multi-scale complex semantic information in the environment monitoring video can be more comprehensively captured, the feature representation can be enhanced, and more accurate and comprehensive data support can be provided for the subsequent environment state type recognition.

[0038] Figure 4 The present invention is a flowchart of inputting each environmental state semantic feature map in the sequence of environmental state semantic feature maps into a feature excitation-squeezing module including a feature multi-scale perception enhancement layer and a multi-scale semantic feature extraction layer to obtain a sequence of environmental state semantic enhancement feature vectors in the digital asset management method based on the environmental protection field according to an embodiment of the present application. Figure 4 As shown, the step S130 includes: S131, performing feature excitation on each environmental state semantic feature map in the sequence of environmental state semantic feature maps through the feature multi-scale perception enhancement layer of the feature excitation-squeezing module to obtain a sequence of environmental state perception enhancement semantic feature maps; S132, performing feature compression on each environmental state perception enhancement semantic feature map in the sequence of environmental state perception enhancement semantic feature maps through the multi-scale semantic feature extraction layer of the feature excitation-squeezing module to obtain a sequence of environmental state semantic enhancement feature vectors.

[0039] Figure 5The present invention is a flowchart of a method for managing digital assets in the field of environmental protection according to an embodiment of the present application, wherein each environmental state semantic feature graph in the sequence of environmental state semantic feature graphs is subjected to feature excitation by the feature multi-scale perception enhancement layer of the feature excitation-squeezing module to obtain a sequence of environmental state perception enhanced semantic feature graphs. Figure 5 As shown, the step S131 includes: S1311, inputting the environment state semantic feature map into the low-frequency detection branch of the feature multi-scale perception enhancement layer to obtain the environment state local low-frequency activation semantic feature vector, wherein the low-frequency detection branch includes a first point convolution layer, a first global mean pooling layer and a first non-linear activation layer; S1312, inputting the environment state semantic feature map into the simplified compression branch of the feature multi-scale perception enhancement layer to obtain the environment state channel compression semantic feature map, wherein the simplified compression branch includes a second point convolution layer; S1313, inputting the environment state semantic feature map into the global space expansion perception branch of the feature multi-scale perception enhancement layer to obtain the environment state receptive field expansion global activation semantic feature matrix, wherein the global space expansion perception branch includes a hole convolution layer, a third point convolution layer and a second non-linear activation layer; S1314, performing feature superposition and integration on the environment state local low-frequency activation semantic feature vector, the environment state channel compression semantic feature map and the environment state receptive field expansion global activation semantic feature matrix to obtain the environment state perception enhancement semantic feature map.

[0040] More specifically, the step S1314 includes: performing positional point multiplication on the environmental state receptive field expansion global activation semantic feature matrix and each feature matrix along the channel dimension of the environmental state channel compression semantic feature map to obtain a channel modulated environmental state global semantic feature map; using each eigenvalue in the environmental state local low-frequency activation semantic feature vector as a weight, performing point multiplication on each feature matrix along the channel dimension of the environmental state channel compression semantic feature map to obtain a channel modulated environmental state local low-frequency semantic feature map; adding the channel modulated environmental state global semantic feature map and the channel modulated environmental state local low-frequency semantic feature map by position to obtain an environmental state multi-scale semantic fusion feature map; performing atrous convolution coding on the environmental state multi-scale semantic fusion feature map to obtain the environmental state perception enhanced semantic feature map.

[0041] That is, the step S131 includes: processing the environment state semantic feature map with the following multi-scale perception enhancement formula to obtain the environment state perception enhancement semantic feature map; wherein the multi-scale perception enhancement formula is:

[0042] Where X is the semantic feature map of the environment state, C1 is a 1×1 point convolution, and C3,3 and C 3,2 are 3×3 dilated convolutions with 3 and 2 dilated numbers, δ represents the sigmoid activation function, and Avg(·) represents the global mean pooling process. represents positional multiplication, represents addition by position, and Y is the environmental state perception enhanced semantic feature map.

[0043] Figure 6 The present invention is a flowchart of a method for managing digital assets in the field of environmental protection according to an embodiment of the present application, wherein each environmental state perception enhancement semantic feature graph in the sequence of environmental state perception enhancement semantic feature graphs is subjected to feature extraction by the multi-scale semantic feature extraction layer of the feature excitation-extraction module to obtain a sequence of environmental state semantic enhancement feature vectors. Figure 6 As shown, the step S132 includes: S1321, upsampling the environmental state perception enhanced semantic feature map to obtain an upsampled environmental state perception enhanced semantic feature map; S1322, passing the upsampled environmental state perception enhanced semantic feature map through the global semantic feature extraction branch of the multi-scale semantic feature extraction layer to obtain an environmental state perception enhanced global semantic feature vector, wherein the global semantic feature extraction module includes a second global mean pooling layer, a fourth point convolution layer, a first batch normalization layer and a third non-linear activation layer; S1323, passing the upsampled environmental state perception enhanced semantic feature map through the local semantic feature extraction branch of the multi-scale semantic feature extraction layer to obtain an environmental state perception enhanced local semantic feature vector, wherein the local semantic feature extraction module includes a fifth point convolution layer, a second batch normalization layer and a fourth non-linear activation layer; S1324, fusing the environmental state perception enhanced global semantic feature vector and the environmental state perception enhanced local semantic feature vector to obtain the environmental state semantic enhancement feature vector.

[0044] In the above-mentioned digital asset management method based on the field of environmental protection, the step S140 inputs the sequence of the environmental state semantic enhancement feature vectors into a contextual semantic association enhancement module based on saliency-globality to obtain a sequence of associated enhanced environmental state semantic feature vectors. It should be understood that since the environmental monitoring video has continuity and dynamics, there is a temporal and spatial correlation between each key frame therein. Therefore, in order to more accurately identify and understand the changes in environmental states, it is necessary to further consider the contextual association information between each environmental monitoring key frame. Based on this, in the technical solution of the present application, a contextual semantic association enhancement module based on saliency-globality is introduced to process the sequence of the environmental state semantic enhancement feature vectors, so as to fully explore and utilize the contextual association information between the environmental monitoring key frames. Specifically, the context semantic association enhancement module extracts the significant features and global features of the sequence of the environmental state semantic enhancement feature vector, and performs convolution, activation and other operations on the significant features and global features to mine the multi-dimensional context semantic associations between each environmental monitoring key frame, thereby calculating the attention weight of the sequence of the environmental state semantic enhancement feature vector, and performing context semantic association enhancement based on the attention mechanism on the sequence of the environmental state semantic enhancement feature vector. Through the introduction of context association information, the spatiotemporal correlation between each environmental monitoring key frame can be fully captured and expressed, and the continuous dynamic changes in the environmental monitoring video, such as biological migration, meteorological patterns, etc., can be more accurately identified and understood, so as to more accurately classify and manage the environmental state.

[0045] Figure 7 The present invention is a flowchart of inputting the sequence of environmental state semantic enhancement feature vectors into a contextual semantic association enhancement module based on saliency-globality in the digital asset management method based on the environmental protection field according to an embodiment of the present application to obtain a sequence of association-enhanced environmental state semantic feature vectors. Figure 7As shown, the step S140 includes: S141, calculating the maximum value of each environmental state semantic enhancement feature vector in the sequence of the environmental state semantic enhancement feature vector to obtain an environmental state semantic salient feature vector composed of each of the maximum values; S142, calculating the average value of each environmental state semantic enhancement feature vector in the sequence of the environmental state semantic enhancement feature vector to obtain an environmental state semantic global feature vector composed of each of the average values; S143, performing one-dimensional convolution coding, nonlinear activation and linear transformation on the environmental state semantic enhancement feature vector and the environmental state semantic global feature vector to obtain an environmental state semantic enhancement feature context association activation feature vector and an environmental state semantic global feature context association activation feature vector; S144, based on the environmental state semantic enhancement feature context association activation feature vector and the environmental state semantic global feature context association activation feature vector, performing semantic association enhancement on the sequence of the environmental state semantic enhancement feature vectors to obtain the sequence of the association enhanced environmental state semantic feature vectors.

[0046] More specifically, the step S144 includes: calculating the position-by-position addition of the environmental state semantic salient feature context-associated activation feature vector and the environmental state semantic global feature context-associated activation feature vector to obtain the environmental state salient-global feature context-associated activation feature vector; performing nonlinear activation based on the tanh activation function and nonlinear activation based on the sigmoid activation function on the environmental state salient-global feature context-associated activation feature vector to obtain a first activation result and a second activation result, and performing position-by-position point multiplication on the first activation result and the second activation result to obtain an environmental state salient-global feature context-associated weighted feature vector; using the feature values ​​of each position in the environmental state salient-global feature context-associated weighted feature vector as weights, weighting the sequence of the environmental state semantic enhancement feature vectors vector by vector, and then adding and fusing them with the sequence of the environmental state semantic enhancement feature vectors to obtain the sequence of the association-enhanced environmental state semantic feature vectors.

[0047] That is, the step S140 includes: processing the sequence of the environmental state semantic enhancement feature vectors with the following contextual semantic enhancement formula to obtain the sequence of the associated enhanced environmental state semantic feature vectors, wherein the contextual semantic enhancement formula is:

[0048] H′=Avg(H)

[0049] H"=Max(H)

[0050]

[0051] V=H⊙(tanh(V1+V2)⊙σ(V1+V2))+H

[0052] Wherein, H is the sequence of the semantic enhancement feature vectors of the environment state, H′ is the global semantic feature vector of the environment state, H″ is the semantic salient feature vector of the environment state, Avg represents the average value, Max represents the maximum value, Conv represents the one-dimensional convolution operation, ReLU represents the ReLU activation function, and are respectively the first weight matrix and the second weight matrix for linear transformation, V1 and V2 are respectively the context-associated activation feature vector of the environmental state semantic global feature and the context-associated activation feature vector of the environmental state semantic salient feature, + represents the sum of corresponding elements, σ ​​represents the Sigmoid function, tanh represents the tanh activation function, ⊙ represents the dot product, and V is the sequence of the association-reinforced environmental state semantic feature vectors.

[0053] In the above-mentioned digital asset management method based on the field of environmental protection, the step S150 automatically archives the digital assets imported by the user based on the sequence of the associated enhanced environmental state semantic feature vectors. Figure 8 The flowchart is a method for managing digital assets in the field of environmental protection according to an embodiment of the present application, which automatically archives the digital assets imported by the user based on the sequence of the associated enhanced environmental state semantic feature vectors. Figure 8 As shown, the step S150 includes: S151, after cascading the sequence of the associated enhanced environmental state semantic feature vectors into a full-time domain environmental state semantic association feature vector, the full-time domain environmental state semantic association feature vector is input into a classifier-based digital asset classifier to obtain a classification result, and the classification result is used to represent the type label to which the digital asset belongs; S152, based on the classification result, automatically archiving the digital assets imported by the user.

[0054] Specifically, in step S151, after the sequence of the associated enhanced environment state semantic feature vectors is cascaded into a full-time environment state semantic associated feature vector, the full-time environment state semantic associated feature vector is input into a classifier-based digital asset classifier to obtain a classification result, and the classification result is used to represent the type label to which the digital asset belongs. That is, by cascading the sequence of the associated enhanced environment state semantic feature vectors to integrate the overall video information, the full-time environment state feature representation is obtained, thereby obtaining the full-time environment state semantic associated feature vector. Then, the full-time environment state semantic associated feature vector is subjected to feature pattern recognition and classification processing by a classifier-based digital asset classifier. Specifically, the classifier has undergone a large amount of training and optimization, and can map the full-time environment state semantic associated feature vector to the corresponding category according to the preset classification label, such as "good environment", "mild pollution", "heavy pollution", etc.

[0055] In particular, in the technical solution of the present application, each environmental state semantic enhancement feature vector in the sequence of environmental state semantic enhancement feature vectors respectively represents the image semantic enhancement features of each environmental monitoring key frame in the sequence of environmental monitoring key frames. When the sequence of environmental state semantic enhancement feature vectors is input into the contextual semantic association enhancement module based on saliency-globality, the contextual semantic association enhancement module based on saliency-globality utilizes global features and prominent features to enhance contextual attention to improve the feature utilization capability of key features, but this will also aggravate the feature distribution differences between each associated enhanced environmental state semantic feature vector in the sequence of associated enhanced environmental state semantic feature vectors, thereby further making the overall feature distribution of the full-time domain environmental state semantic association feature vector obtained by cascading have the problem of insufficient feature aggregation, thereby affecting its classification convergence efficiency through the classifier, that is, affecting the efficiency of classification training and the accuracy of classification results.

[0056] Based on this, in a preferred embodiment of the present application, the full-time-domain environmental state semantic association feature vector is passed through a classifier-based digital asset classifier to obtain a classification result, which specifically includes the following steps: after dot-multiplying the full-time-domain environmental state semantic association feature vector with the length of the full-time-domain environmental state semantic association feature vector, subtracting the dot-value from the first norm of the full-time-domain environmental state semantic association feature vector to obtain a full-time-domain environmental state semantic association full semantic representation vector; after dot-multiplying the full-time-domain environmental state semantic association feature vector with the square root of the length of the full-time-domain environmental state semantic association feature vector, subtracting the dot-value from the second norm of the full-time-domain environmental state semantic association feature vector to obtain a full-time-domain environmental state semantic association semi-semantic representation vector; respectively calculating the full-time-domain environmental state semantic association full semantic representation vector and the full-time-domain environmental state semantic association vector. The square root of the absolute value of each eigenvalue of the semi-semantic representation vector is connected to obtain a full-time-domain environmental state semantic association full semantic modulation vector and a full-time-domain environmental state semantic association semi-semantic modulation vector; the logarithm with base 2 of each eigenvalue of the full-time-domain environmental state semantic association full semantic modulation vector and the full-time-domain environmental state semantic association semi-semantic modulation vector are calculated respectively to obtain a full-time-domain environmental state semantic association full semantic information vector and a full-time-domain environmental state semantic association semi-semantic information vector; the weighted sum of the full-time-domain environmental state semantic association full semantic information vector and the full-time-domain environmental state semantic association semi-semantic information vector is calculated with a weighting coefficient as a hyperparameter to obtain an optimized full-time-domain environmental state semantic association feature vector; the optimized full-time-domain environmental state semantic association feature vector is passed through the classifier-based digital asset classifier to obtain the classification result.

[0057] Accordingly, in this preferred embodiment, the full-time-domain environmental state semantic association feature vector is taken as a feature set, wherein the change semantic representation is based on the feature value of the full-time-domain environmental state semantic association feature vector. In order to dynamically aggregate the semantic set as a whole composed of different change semantics of the full-time-domain environmental state semantic association feature vector without ignoring the individual semantic change information, the individual features of the full-time-domain environmental state semantic association feature vector and the collective expression of the vector scale of the full-time-domain environmental state semantic association feature vector are used as full-width and half-width feature representations, and the low-rank norms of different dimensions of the overall semantics of the feature set of the full-time-domain environmental state semantic association feature vector are used as phases through negative correlation calculation and amplitude-phase information representation modulation is performed to dynamically adjust the change relationship of the semantic content of the full-time-domain environmental state semantic association feature vector, thereby improving the aggregation of the semantic information expression of the overall feature set of the full-time-domain environmental state semantic association feature vector, and improving the classification convergence efficiency of the full-time-domain environmental state semantic association feature vector when it is classified by a classifier-based digital asset classifier.

[0058] Specifically, the step S152 automatically archives the digital assets imported by the user based on the classification result. That is, the digital assets are automatically classified and stored according to the type tags to which they belong based on the classification result, avoiding the tediousness and time-consuming manual classification, greatly improving the management efficiency of digital assets, thereby realizing the intelligent, efficient and accurate management of digital assets in the field of environmental protection, so that relevant personnel can access and use digital assets more conveniently, so as to timely discover potential environmental problems and take corresponding measures.

[0059] In summary, the digital asset management method based on the environmental protection field according to the embodiment of the present application is explained, which uses machine vision technology based on deep learning to perform semantic analysis on the environmental monitoring video imported by the user, extracts the image semantic features of each key frame in the environmental monitoring video, and strengthens its contextual semantic association to mine the full-time domain environmental status information, thereby intelligently identifying the type of environmental status and realizing automatic archiving and management of digital assets in the environmental protection field. In this way, the management efficiency of environmental data can be improved, the need for manual operation can be reduced, and the accuracy and real-time performance of data management can be improved.

[0060] In another embodiment of the present application, an application of the above-mentioned digital asset management method based on the field of environmental protection in environmental protection is provided. It can be understood that in the field of environmental protection, the use of the above-mentioned digital asset management method can realize the automatic archiving and management of environmental monitoring data, thereby greatly simplifying the process of environmental protection work and improving work efficiency.

[0061] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0062] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0063] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0064] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0065] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A digital asset management method based on the field of environmental protection, characterized in that: include: Acquire a digital asset imported by a user, wherein the digital asset is an environmental monitoring video; Extracting environmental state features of the environmental monitoring video to obtain a sequence of environmental state semantic feature graphs; Inputting each of the environmental state semantic feature maps in the sequence of environmental state semantic feature maps into a feature excitation-squeezing module including a feature multi-scale perception enhancement layer and a multi-scale semantic feature extraction layer to obtain a sequence of environmental state semantic enhancement feature vectors; Inputting the sequence of the environment state semantic enhancement feature vectors into a contextual semantic association enhancement module based on saliency-globality to obtain a sequence of association enhanced environment state semantic feature vectors; Based on the sequence of the associated enhanced environment state semantic feature vectors, the digital assets imported by the user are automatically archived.

2. The digital asset management method based on the environmental protection field according to claim 1 is characterized in that: Extracting environmental state features of the environmental monitoring video to obtain a sequence of environmental state semantic feature graphs includes: Performing key frame sampling on the environment monitoring video to obtain a sequence of environment monitoring key frames; Each environment monitoring key frame in the sequence of environment monitoring key frames is passed through an environment state feature extractor based on a hole convolutional neural network model to obtain a sequence of environment state semantic feature maps.

3. The digital asset management method based on the environmental protection field according to claim 2 is characterized in that: Inputting each of the environmental state semantic feature maps in the sequence of environmental state semantic feature maps into a feature excitation-squeezing module including a feature multi-scale perception enhancement layer and a multi-scale semantic feature extraction layer to obtain a sequence of environmental state semantic enhancement feature vectors, including: Each of the environmental state semantic feature graphs in the sequence of environmental state semantic feature graphs is subjected to feature excitation by the feature multi-scale perception enhancement layer of the feature excitation-squeezing module to obtain a sequence of environmental state perception enhanced semantic feature graphs; Each environmental state perception enhanced semantic feature map in the sequence of environmental state perception enhanced semantic feature maps is subjected to feature compression through the multi-scale semantic feature extraction layer of the feature excitation-compression module to obtain a sequence of environmental state semantic enhanced feature vectors.

4. The digital asset management method based on the environmental protection field according to claim 3 is characterized in that: The process of performing feature excitation on each of the environment state semantic feature graphs in the sequence of environment state semantic feature graphs through the feature multi-scale perception enhancement layer of the feature excitation-squeezing module to obtain a sequence of environment state perception enhancement semantic feature graphs includes: Inputting the environment state semantic feature map into the low-frequency detection branch of the feature multi-scale perception enhancement layer to obtain a local low-frequency activation semantic feature vector of the environment state, wherein the low-frequency detection branch includes a first point convolution layer, a first global mean pooling layer, and a first nonlinear activation layer; Inputting the environment state semantic feature map into the simplification and compression branch of the feature multi-scale perception enhancement layer to obtain an environment state channel compressed semantic feature map, wherein the simplification and compression branch includes a second point convolution layer; Inputting the environment state semantic feature map into the global space expansion perception branch of the feature multi-scale perception enhancement layer to obtain an environment state receptive field expansion global activation semantic feature matrix, wherein the global space expansion perception branch includes a hole convolution layer, a third point convolution layer and a second nonlinear activation layer; The local low-frequency activation semantic feature vector of the environmental state, the channel compression semantic feature map of the environmental state and the global activation semantic feature matrix of the environmental state receptive field expansion are subjected to feature superposition and integration to obtain the environmental state perception enhancement semantic feature map.

5. The digital asset management method based on the environmental protection field according to claim 4 is characterized in that: The environmental state local low-frequency activation semantic feature vector, the environmental state channel compression semantic feature map and the environmental state receptive field expansion global activation semantic feature matrix are subjected to feature superposition and integration to obtain the environmental state perception enhancement semantic feature map, including: Multiply the environmental state receptive field expansion global activation semantic feature matrix by each feature matrix along the channel dimension of the environmental state channel compression semantic feature map by position point to obtain a channel modulated environmental state global semantic feature map; Taking each eigenvalue in the local low-frequency activation semantic feature vector of the environmental state as a weight, performing point multiplication on each feature matrix along the channel dimension of the channel compression semantic feature map of the environmental state to obtain a channel-modulated local low-frequency semantic feature map of the environmental state; Adding the channel modulation environment state global semantic feature map and the channel modulation environment state local low-frequency semantic feature map according to position to obtain an environment state multi-scale semantic fusion feature map; The multi-scale semantic fusion feature map of the environmental state is subjected to hole convolution coding to obtain the environmental state perception enhanced semantic feature map.

6. The digital asset management method based on the environmental protection field according to claim 5 is characterized in that: The process of subjecting each of the environmental state perception enhanced semantic feature maps in the sequence of environmental state perception enhanced semantic feature maps to feature extraction through the multi-scale semantic feature extraction layer of the feature excitation-extraction module to obtain a sequence of environmental state semantic enhanced feature vectors includes: Upsampling the environmental state perception enhanced semantic feature map to obtain an upsampled environmental state perception enhanced semantic feature map; Passing the upsampled environment state perception enhanced semantic feature map through the global semantic feature extraction branch of the multi-scale semantic feature extraction layer to obtain an environment state perception enhanced global semantic feature vector, wherein the global semantic feature extraction module includes a second global mean pooling layer, a fourth point convolution layer, a first batch normalization layer and a third nonlinear activation layer; Passing the upsampled environment state perception enhanced semantic feature map through the local semantic feature extraction branch of the multi-scale semantic feature extraction layer to obtain an environment state perception enhanced local semantic feature vector, wherein the local semantic feature extraction module includes a fifth point convolution layer, a second batch normalization layer and a fourth nonlinear activation layer; The environmental state perception enhanced global semantic feature vector and the environmental state perception enhanced local semantic feature vector are fused to obtain the environmental state semantic enhanced feature vector.

7. The digital asset management method based on the environmental protection field according to claim 6 is characterized in that: Inputting the sequence of the environment state semantic enhancement feature vectors into a context semantic association enhancement module based on saliency-globality to obtain a sequence of association enhanced environment state semantic feature vectors, including: Calculating the maximum value of each environmental state semantic enhancement feature vector in the sequence of environmental state semantic enhancement feature vectors to obtain an environmental state semantic salient feature vector composed of each of the maximum values; Calculating the average value of each environmental state semantic enhancement feature vector in the sequence of environmental state semantic enhancement feature vectors to obtain an environmental state semantic global feature vector composed of each of the average values; Performing one-dimensional convolution coding, nonlinear activation and linear transformation on the environment state semantic salient feature vector and the environment state semantic global feature vector to obtain an environment state semantic salient feature context associated activation feature vector and an environment state semantic global feature context associated activation feature vector; Based on the environment state semantic salient feature context association activation feature vector and the environment state semantic global feature context association activation feature vector, the sequence of the environment state semantic reinforcement feature vectors is semantically reinforced to obtain the sequence of the association reinforced environment state semantic feature vectors.

8. The digital asset management method based on the environmental protection field according to claim 7 is characterized in that: Based on the environment state semantic salient feature context associated activation feature vector and the environment state semantic global feature context associated activation feature vector, the sequence of the environment state semantic enhanced feature vectors is semantically enhanced to obtain the sequence of the associated enhanced environment state semantic feature vectors, including: Calculate the positional addition of the environment state semantic salient feature context associated activation feature vector and the environment state semantic global feature context associated activation feature vector to obtain the environment state salient-global feature context associated activation feature vector; The environmental state prominence-global feature context association activation feature vector is respectively subjected to nonlinear activation based on a tanh activation function and nonlinear activation based on a sigmoid activation function to obtain a first activation result and a second activation result, and the first activation result and the second activation result are multiplied by position points to obtain an environmental state prominence-global feature context association weight feature vector; The feature values ​​of each position in the environmental state salient-global feature context associated weight feature vector are used as weights, the sequence of environmental state semantic enhancement feature vectors is weighted vector by vector, and then added and fused with the sequence of environmental state semantic enhancement feature vectors to obtain the sequence of associated enhanced environmental state semantic feature vectors.

9. The digital asset management method based on the environmental protection field according to claim 8 is characterized in that: Automatically archiving the digital assets imported by the user based on the sequence of the associated enhanced environment state semantic feature vectors includes: After cascading the sequence of the association-enhanced environment state semantic feature vectors into a full-time environment state semantic association feature vector, the full-time environment state semantic association feature vector is input into a classifier-based digital asset classifier to obtain a classification result, wherein the classification result is used to indicate the type label to which the digital asset belongs; Based on the classification result, the digital assets imported by the user are automatically archived.

10. An application of the digital asset management method according to any one of claims 1 to 9 in environmental protection.

Citation Information

Patent Citations

  • Structured filing system for medical data

    CN116776872A

  • Intelligent filing control method for electric power industry bills

    CN116778513A

  • Video understanding method and device, electronic equipment and storage medium

    CN116939287A

  • Method for improving saline-alkali soil based on fly ash

    CN118504793A

  • Wind power generation equipment immersive simulation practical training system based on virtual reality

    CN118535021A