Evidence element identification method for ecological and environmental protection cases and related equipment

Abstract

The present invention provides a method and related equipment for identifying evidence elements in ecological and environmental protection cases. By adaptively sampling the physical information data and digital information data related to the target ecological and environmental protection cases, key evidence data is obtained, which greatly improves the efficiency of data acquisition; the key evidence data is input into the attention distribution model for feature extraction to obtain the key evidence element feature vector; the key evidence element feature vector is input into the evidence element recognition model for recognition to obtain the evidence element recognition result of the target ecological and environmental protection case, which can more accurately screen out the characteristics of the key evidence elements, improve the accuracy and sufficiency of the evidence, help resolve legal disputes over the ecological environment, and help achieve judicial justice in ecological and environmental protection cases.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and judicial technology, and in particular to a method for identifying evidence elements in ecological and environmental protection cases and related equipment. Background Art

[0002] In recent years, the imbalance of the entire ecosystem caused by global climate anomalies, severe soil and water pollution, etc., has seriously threatened human survival and development. As people's awareness of environmental protection becomes stronger and stronger, illegal acts such as environmental pollution have attracted more and more attention from the public. As a global issue, ecological and environmental protection has received increasing attention from governments around the world. Ecological and environmental protection cases have more complex case elements. Their cases are generally intricate and the causal relationships between different factors are interrelated. Therefore, the accuracy and sufficiency of evidence are crucial to winning the case. At present, the traditional methods of obtaining evidence mainly include the following:

[0003] (1) Evidence collection method based on manual labor: This method relies on manual collection and screening of evidence data, which is cumbersome and time-consuming, and has shortcomings in accuracy and efficiency;

[0004] (2) Automatic image analysis technology based on machine learning: This technology uses computer vision and machine learning algorithms to automatically analyze and identify collected images. However, it also has problems with accuracy and recognition rate, and cannot cover a wider range of evidence.

[0005] (3) Monitoring and collection systems or tools: Some existing monitoring and collection systems or tools, such as business supervision systems and intelligent hardware devices, can collect relevant information and evidence materials for environmental protection cases, but they still require manual analysis and judgment of the data, and cannot achieve automatic classification and analysis of evidence. Summary of the Invention

[0006] The present invention provides a method for identifying evidence elements in ecological and environmental protection cases and related equipment, the purpose of which is to improve the efficiency and accuracy of obtaining evidence in ecological and environmental protection cases.

[0007] To achieve the above objectives, the present invention provides a method for identifying evidence elements in ecological and environmental protection cases, comprising:

[0008] Step 1: Collect physical and digital information data related to the target ecological and environmental protection case. Physical information data includes environmental monitoring data related to the target ecological and environmental protection case acquired using monitoring equipment. Digital information data includes text, image, and video data related to the target ecological and environmental protection case.

[0009] Step 2: Establish an adaptive sampling model including a sensitivity analysis module based on the Markov Chain Monte Carlo algorithm. Input the physical information data and digital information data into the adaptive sampling model. The sensitivity analysis module analyzes the importance and sensitivity of the physical information data and digital information data. Adaptive sampling is then performed on the physical information data and digital information data based on the importance and sensitivity to obtain key evidence data.

[0010] Step 3: The key evidence data input is introduced into the attention distribution model with a self-supervised learning mechanism for feature extraction to obtain the spatial features of the key evidence data, the global spatial relationships and temporal relationships in the key evidence data, and the spatial features of the key evidence data and the global spatial relationships and temporal relationships in the key evidence data are integrated to obtain the key evidence element feature vector; the local spatial features of the key evidence data include the texture features, shape features, pollution features and structural features of the ecological environment; the global spatial relationships and temporal relationships are the relationships and temporal relationships of the contextual information in the key evidence data;

[0011] Step 4: Input the key evidence element feature vector into the evidence element recognition model for recognition, and obtain the evidence element recognition results of the target ecological and environmental protection case;

[0012] The attention distribution model includes: an embedding module, a long short-term memory module, a convolutional neural network module and a fusion module; the input end of the embedding module is connected to the input end of the long short-term memory module, the output end of the long short-term memory module is connected to the input end of the convolutional neural network module, the output end of the convolutional neural network module and the output end of the long short-term memory module are both connected to the input end of the fusion module, and the output end of the fusion module is connected to the input end of the evidence element recognition model.

[0013] Furthermore, before step 2, it also includes:

[0014] Preprocessing the physical information data and the digital information data to obtain preprocessed physical information data and digital information data;

[0015] The preprocessed physical information data and the preprocessed digital information data are input into the adaptive sampling model for adaptive sampling.

[0016] More specifically, the adaptive sampling model includes:

[0017] The Markov chain Monte Carlo module, sensitivity analysis module and adaptive sampling module are connected in sequence;

[0018] Inputting the preprocessed physical information data and the preprocessed digital information data into the Markov chain Monte Carlo algorithm for sampling to obtain physical information data and digital information data that meet the preset probability distribution;

[0019] Input the physical information data and digital information data that meet the preset probability into the sensitivity analysis module, and analyze the importance and sensitivity of the physical information data and digital information data that meet the preset probability distribution through the sensitivity analysis module to obtain the importance and sensitivity of the physical information data and digital information data as well as sampling weights;

[0020] The importance and sensitivity of physical information data and digital information data as well as the sampling weights are input into the adaptive sampling module, and the sampling probabilities of physical information data and digital information data are adaptively adjusted through the adaptive sampling module to obtain key evidence data.

[0021] Furthermore, the attention distribution model also includes:

[0022] Convolution module and pooling module;

[0023] The input of the convolution module is connected to the output of the long short-term memory module, the output of the convolution layer is connected to the input of the pooling module, and the output of the pooling module is connected to the input of the convolutional neural network module.

[0024] More specifically, step 3 includes:

[0025] Input the key evidence data into the embedding module of the attention distribution model for conversion to obtain the vector of the key evidence data;

[0026] By extracting features from vectors in the long short-term memory module, sequence data features are obtained;

[0027] The sequence data features are input into the convolution module for the first feature extraction to obtain local spatial features and generate feature maps;

[0028] The feature map is input into the pooling module for the second feature extraction to obtain structural features;

[0029] The local spatial features and structural features are input into the convolutional neural network module for feature extraction to obtain the global spatial and temporal relationships in the key evidence data;

[0030] The local spatial features, structural features and the global spatial and temporal relationships in the key evidence data are input into the fusion module for weighted fusion to obtain the key evidence element feature vector.

[0031] Furthermore, before inputting the key evidence element feature vector into the evidence element recognition model for recognition, the following steps are also included:

[0032] Classify the key evidence element feature vectors through a classifier to obtain text feature vectors, image feature vectors, and video feature vectors;

[0033] The text feature vector, image feature vector and video feature vector are input into the evidence element recognition model for recognition.

[0034] Specifically, the evidence element identification model includes:

[0035] A first conversion module, a second conversion module, a third conversion module, a multi-task learning module, and an attention module;

[0036] The input end of the first conversion module, the input end of the second conversion module, and the input end of the third conversion module are all connected to the output end of the classifier, the output end of the first conversion module, the output end of the second conversion module, and the output end of the third conversion module are all connected to the input end of the multi-task learning module, and the output end of the multi-task learning module is connected to the input end of the attention module;

[0037] Inputting the text feature vector, the image feature vector and the video feature vector into the first conversion module, the second conversion module and the third conversion module respectively for conversion to obtain text data, image data and key frame images;

[0038] Inputting text data, image data and key frame images into a multi-task learning module for multi-task learning to obtain learned text data, learned image data and learned key frame images;

[0039] The learned text data, learned image data, and learned key frame images are input into the attention module for recognition to obtain the evidence element recognition results of the target ecological and environmental protection case. The evidence element recognition results include: the category prediction probability distribution of the learned text data, the learned image data and category labels, and the key frames of the learned key frame images.

[0040] Furthermore, after step 4, the following steps are also included:

[0041] A graph neural network is used to extract the relationship between each evidence element in the evidence element recognition results to obtain the correlation relationship between each evidence element;

[0042] Classify and sort out the correlations between the various elements of evidence to obtain classification results;

[0043] Based on the classification results, the association rule learning algorithm is used to find the frequent patterns between the various evidence elements and obtain the key events;

[0044] Based on the correlation and key events, we can obtain the doubts and proof directions of the target ecological and environmental protection cases;

[0045] Extract related information and clues based on doubts and their proof direction to build a chain of evidence for target ecological and environmental protection cases.

[0046] The present invention also provides a computer storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, a method for identifying evidence elements in ecological and environmental protection cases is implemented.

[0047] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, a method for identifying evidence elements in ecological and environmental protection cases is implemented.

[0048] The above solution of the present invention has the following beneficial effects:

[0049] Compared with the existing technology, the present invention uses an adaptive sampling model to adaptively sample physical information data and digital information data related to target ecological and environmental protection cases to obtain key evidence data, thereby greatly improving the efficiency of data acquisition; the key evidence data is input into the attention distribution model for feature extraction to obtain the key evidence element feature vector; the key evidence element feature vector is input into the evidence element recognition model for recognition to obtain the evidence element recognition result of the target ecological and environmental protection case, which can more accurately screen out the characteristics of the key evidence elements, improve the accuracy and sufficiency of the evidence, help resolve legal disputes over the ecological environment, and help achieve judicial justice in ecological and environmental protection cases.

[0050] Other beneficial effects of the present invention will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A schematic diagram of a flow chart of an embodiment of the present invention;

[0052] Figure 2 Schematic diagram of the structure of the adaptive sampling model in an embodiment of the present invention;

[0053] Figure 3 Schematic diagram of the structure of the attention distribution model in an embodiment of the present invention;

[0054] Figure 4 Schematic diagram of the structure of the evidence element identification model in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] To make the technical problems, technical solutions, and advantages to be solved by the present invention more clear, the following is a detailed description with reference to the accompanying drawings and specific embodiments. It is obvious that the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0056] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0057] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood broadly. For example, they may refer to a locking connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0058] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0059] In response to existing problems, the present invention provides a method for identifying evidence elements in ecological and environmental protection cases and related equipment.

[0060] like Figure 1 As shown, an embodiment of the present invention provides a method for identifying evidence elements in ecological and environmental protection cases, including:

[0061] Step 1: Collect physical and digital information data related to the target ecological and environmental protection case. Physical information data includes environmental monitoring data related to the target ecological and environmental protection case acquired using monitoring equipment. Digital information data includes text, image, and video data related to the target ecological and environmental protection case.

[0062] Specifically, the environmental monitoring data collected by the embodiment of the present invention is monitoring data for substances such as water, air and soil, which can be obtained by physical monitoring means such as sensors or other monitoring equipment; text information data is text information captured from relevant websites, forums, Weibo and other social platforms using web crawler technology, such as news reports, announcements, laws and regulations, etc., and converted into digital information data according to data standards; image information data is based on computer vision technology, and the improved R-CNN algorithm automatically identifies image information related to the target ecological and environmental protection case, such as photos of pollution around construction sites or factories, photos of biological organisms in wildlife reserves, etc., and records it as digital information data; the improved R-CNN algorithm uses YOLO to improve the R-CNN algorithm: the R-FCN candidate region extraction technology is used to replace the selected Selective search, for feature images of different scales, a feature fusion strategy is adopted to perform weighted fusion of feature images of different scales to enhance the feature expression capability and make the entire algorithm faster. The Anchor in YOLO is changed to the candidate region extraction method in R-CNN, and the feature extraction operation is replaced by the convolutional neural network feature extraction in R-CNN, and the category month, position prediction and confidence prediction operations are added; distributed training is used to improve target detection accuracy; at the same time, a multi-scale fusion strategy is used to fuse features at different levels to improve detection accuracy and stability; video information data is continuously collected from real-time monitoring cameras in public places, factories, offices, etc. through automated video processing technology, and key frames are sampled and extracted and recorded as digital information data.

[0063] Step 2: Establish an adaptive sampling model including a sensitivity analysis module based on the Markov Chain Monte Carlo algorithm. Input the physical information data and digital information data into the adaptive sampling model. The sensitivity analysis module analyzes the importance and sensitivity of the physical information data and digital information data. Adaptive sampling is then performed on the physical information data and digital information data based on the importance and sensitivity to obtain key evidence data.

[0064] Step 3: The key evidence data input is introduced into the attention distribution model with a self-supervised learning mechanism for feature extraction to obtain the spatial features of the key evidence data, the global spatial relationships and temporal relationships in the key evidence data, and the spatial features of the key evidence data and the global spatial relationships and temporal relationships in the key evidence data are integrated to obtain the key evidence element feature vector; the local spatial features of the key evidence data include the texture features, shape features, pollution features and structural features of the ecological environment; the global spatial relationships and temporal relationships are the relationships and temporal relationships of the contextual information in the key evidence data;

[0065] Step 4: Input the key evidence element feature vector into the evidence element recognition model for recognition, and obtain the evidence element recognition results of the target ecological and environmental protection case.

[0066] Specifically, before step 2, it also includes:

[0067] Preprocessing the physical information data and the digital information data to obtain preprocessed physical information data and digital information data;

[0068] The preprocessed physical information data and the preprocessed digital information data are input into the adaptive sampling model for adaptive sampling.

[0069] In an embodiment of the present invention, the physical information data and digital information data are first subjected to preprocessing operations such as noise filtering and data cleaning to obtain preprocessed physical information data and digital information data to ensure data quality; then, an adaptive sampling model is used to adaptively sample the preprocessed physical information data and digital information data.

[0070] The basic idea of adaptive sampling is to minimize the redundancy of sample data and retain the most representative data samples based on the data distribution characteristics and the importance and sensitivity of the data, while ensuring sample coverage. This step aims to remove useless information and retain only key evidence data. The following formula can be used for adaptive sampling processing:

[0071]

[0072] Among them, s k represents the probability that data point k is retained, z k represents the eigenvalue of data point k, θ k and σ represent adjustable parameters and standard deviation respectively, f(x,θ f ) represents the feature learning function, which is used to extract key features.

[0073] More specifically, if Figure 2 As shown, the adaptive sampling model includes:

[0074] The Markov chain Monte Carlo module, sensitivity analysis module and adaptive sampling module are connected in sequence;

[0075] Among them, the Markov Chain Monte Carlo module integrates the Markov Chain Monte Carlo algorithm (MCMC), which is used to simulate complex probability distributions and can generate samples that meet specific probability distributions.

[0076] Inputting the preprocessed physical information data and the preprocessed digital information data into the Markov chain Monte Carlo algorithm for sampling to obtain physical information data and digital information data that meet the preset probability distribution;

[0077] The physical information data and digital information data that meet the preset probability are input into the sensitivity analysis module. The sensitivity analysis module analyzes the importance and sensitivity of the physical information data and digital information data that meet the preset probability distribution to obtain the importance and sensitivity of the physical information data and digital information data as well as the sampling weight. The sensitivity analysis module is used to evaluate the importance and sensitivity of the data. Based on pre-defined factors and rules, it performs sensitivity analysis on each data and determines its weight in the sampling process so that these factors can be considered in the sampling process. The weight is quantified and designed in combination with the characteristics of ecological and environmental protection cases and the specific case causes. These factors can be the environmental impact, social impact and legal compliance of the data.

[0078] The importance and sensitivity of physical information data and digital information data, as well as the sampling weights, are input into the adaptive sampling module. The sampling probability of physical information data and digital information data is adaptively adjusted by the adaptive sampling module, so that data with higher sensitivity and importance can be sampled, thereby selecting the most representative key evidence data with the highest recognition rate from the collected data, providing a high-quality data source for subsequent evidence element extraction and identification.

[0079] Specifically, if Figure 3 As shown, the attention distribution model includes: an embedding module, a long short-term memory module, a convolution module, a pooling module, a convolutional neural network module and a fusion module; the input end of the embedding module is connected to the input end of the long short-term memory module, the output end of the long short-term memory module is connected to the input end of the convolution module, the output end of the convolution layer is connected to the input end of the pooling module, the output end of the pooling module is connected to the input end of the convolutional neural network module, the output end of the convolutional neural network module and the output end of the long short-term memory module are both connected to the input end of the fusion module, and the output end of the fusion module is connected to the input end of the evidence element recognition model.

[0080] Specifically, step 3 includes:

[0081] Input the key evidence data into the embedding module of the attention distribution model for conversion to obtain the vector of the key evidence data;

[0082] By extracting features from vectors in the long short-term memory module, sequence data features are obtained;

[0083] The sequence data features are input into the convolution module for the first feature extraction to obtain local spatial features and generate feature maps. Local spatial features mainly refer to the following features:

[0084] (1) Texture features: Different areas of ecological transformation usually have different texture features, such as vegetation cover and soil texture. These texture features can be captured by the convolution module to help identify surface cover, soil type, etc.

[0085] (2) Shape features: Different objects and elements in the ecological environment usually have different shape features, such as the curvature of a river, the shape of a forest, etc. These shape features are captured by the convolution module to help identify different ecological environment elements;

[0086] (3) Pollution characteristics: Pollutants in the ecological environment usually have certain spatial distribution characteristics, such as concentration changes around the pollution source and the range of pollutant diffusion. These pollution characteristics can be extracted through the convolution module to help identify the pollution source and the spread of pollutants.

[0087] The feature map is input into the pooling module for the second feature extraction to obtain the structural features and converted into a probability distribution through the SoftMax layer. The output of the entire vector is processed into a probability distribution so that the sum of the output probabilities of each category is 1;

[0088] Structural characteristics refer to the fact that different elements in an ecological environment usually have certain spatial structural relationships, such as the relationship between lakes and rivers, and the relationship between mountains and forests;

[0089] The local spatial features and structural features are input into the convolutional neural network module for feature extraction to obtain the global spatial and temporal relationships in the key evidence data; the global spatial and temporal relationships capture the contextual information and temporal relationships in the evidence data during the feature extraction process;

[0090] The local spatial features, structural features, and global spatial and temporal relationships in the key evidence data are input into the fusion module for weighted fusion to obtain the key evidence element feature vector, which is expressed as follows:

[0091]

[0092] in, represents the weighted data point k, represents the weight associated with the attention distribution, x k represents the features of data point k.

[0093] Specifically, in order to more accurately extract features closely related to target ecological and environmental protection cases, a self-supervised learning mechanism was introduced in the process of training the attention distribution model. This mechanism can self-guide the model's learning process by predicting the contextual information of the evidence data, thereby improving the accuracy and robustness of feature extraction; based on the above-mentioned attention distribution model with the introduction of a self-supervised learning mechanism, it mainly changes the traditional way of extracting evidence features, which can not only improve the accuracy of feature extraction, but also effectively reduce the model's sensitivity to noise data, giving it stronger generalization ability.

[0094] Specifically, before inputting the key evidence element feature vector into the evidence element recognition model for recognition, it also includes:

[0095] Classify the key evidence element feature vectors through a classifier to obtain text feature vectors, image feature vectors, and video feature vectors;

[0096] The text feature vector, image feature vector and video feature vector are input into the evidence element recognition model for recognition.

[0097] The attention distribution model in the embodiment of the present invention adopts a combination of long short-term memory network LSTM and convolutional neural network CNN. Compared with using only a single network model, the combination of the two network models can better capture the temporal, textual and spatial features in the input key evidence data, thereby improving the performance and accuracy of the model. For scenarios where images and texts are processed simultaneously, an attention distribution model composed of a LSTM-CNN hybrid can be used to process the data, and the output results of LSTM and CNN are fused to provide a comprehensive key evidence element feature vector, which comprehensively considers temporal, textual and spatial information while achieving better results in classification accuracy; the key evidence element feature vector is then input into a classifier for text and image classification, and is trained and optimized. In practical applications, different models and technologies can be used for adjustment and improvement, such as convolutional LSTM models, bidirectional LSTM models, multi-task learning and other methods.

[0098] Compared with traditional methods, it has better feature extraction capabilities. Through automatic learning and optimization of the model, it can better reflect the complexity and nonlinearity of the data. It has the ability to extract time series, text and spatial features and can adaptively adjust model parameters.

[0099] Specifically, if Figure 4 As shown in the figure, the evidence element identification model includes:

[0100] A first conversion module, a second conversion module, a third conversion module, a multi-task learning module, and an attention module;

[0101] The input end of the first conversion module, the input end of the second conversion module, and the input end of the third conversion module are all connected to the output end of the classifier, the output end of the first conversion module, the output end of the second conversion module, and the output end of the third conversion module are all connected to the input end of the multi-task learning module, and the output end of the multi-task learning module is connected to the input end of the attention module;

[0102] Inputting the text feature vector, the image feature vector and the video feature vector into the first conversion module, the second conversion module and the third conversion module respectively for conversion to obtain text data, image data and key frame images;

[0103] Inputting text data, image data and key frame images into a multi-task learning module for multi-task learning to obtain learned text data, learned image data and learned key frame images;

[0104] The learned text data, learned image data and learned key frame images are input into the attention module for recognition. By using different attention mechanisms to process the output text and image data, the evidence element recognition results of the target ecological and environmental protection cases are obtained. The evidence element recognition results include: the category prediction probability distribution of the learned text data, the learned image data and category labels, and the key frames of the learned key frame images.

[0105] In an embodiment of the present invention, a multi-task learning (MTL) module is used to improve performance by utilizing useful information between multiple related learning tasks. Multi-task learning can share the features of three tasks. The output result of text data is a predicted probability distribution belonging to different categories. The output result of image data is usually represented by a single category label, that is, the input image is finally classified into a certain category. The video data is converted into image key frames, so that the model can share information from different tasks. During the training process, soft constraints are used to adjust the correlation relationship between tasks to optimize the total loss function of multiple tasks and improve the generalization ability of the model; the text data uses a self-attention mechanism to identify key evidence elements in the text data, and realizes the summary and adaptive feature selection of sequence information through self-association. The image uses a cross-attention mechanism to capture the interaction between visual features and connect the representation of the target object with other related objects; then the evidence elements are weighted to obtain the key evidence element recognition result, and the weight is calculated. The weight of the key evidence element is increased by weighting. The calculation formula is as follows:

[0106]

[0107] Among them, a represents the weighted evidence element, x iThey represent the weighting coefficient and the characteristics of the evidential elements respectively.

[0108] Compared with existing recognition algorithms, the evidence element recognition model provided by the embodiment of the present invention combines the idea of multi-task learning and the attention mechanism. This method can be customized and optimized for different data types and task requirements, and can achieve better classification accuracy and recognition results, with better generalization and interpretability.

[0109] Specifically, after step 4, it also includes:

[0110] A graph neural network is used to extract the relationship between each evidence element in the evidence element recognition results to obtain the correlation relationship between each evidence element;

[0111] Classify and sort out the correlations between the various elements of evidence to obtain classification results;

[0112] Based on the classification results, the association rule learning algorithm is used to find the frequent patterns between the various evidence elements and obtain the key events;

[0113] Based on the correlation and key events, we can obtain the doubts and proof directions of the target ecological and environmental protection cases;

[0114] Extract related information and clues based on doubts and their proof direction to build a chain of evidence for target ecological and environmental protection cases.

[0115] Graph neural networks are used to extract and classify relationships between evidence elements. Specifically, a graph neural network (GNN) is employed to extract and classify relationships between evidence elements. First, each evidence element is treated as a node in a graph. Graph edges are then constructed based on the relationships between evidence elements. Next, the GNN's powerful node information aggregation capabilities are leveraged to extract and learn contextual information for each node (i.e., evidence element), resulting in a contextual feature vector for each evidence element. Finally, these feature vectors are fed into a multi-layer perceptron (MLP) for evidence element classification. Combined with the experience of judicial experts and precedents, the evidence elements are further screened and evaluated for accuracy to ensure the most representative and credible evidence is mined from the collected data. Based on the classification results, an association rule learning algorithm is used to identify key events. In this step, the evidence elements are first grouped according to their classification results. Within each group, an association rule learning algorithm is then used to discover frequent patterns, identifying patterns that frequently occur across multiple evidence elements. These patterns represent key events requiring attention. Based on key events, a Bayesian network is used to infer the doubts and proof directions of the target ecological and environmental protection case. In this step, a Bayesian network is used for inference. First, the Bayesian network structure is constructed based on the key events. Then, the Bayesian network's conditional probability table is calculated based on the classification results of the evidentiary elements. Next, the Bayesian network is used for inference to determine the doubts and proof directions of the target ecological and environmental protection case. This method, which combines graph neural networks, association rule learning algorithms, and Bayesian networks, effectively extracts key information from the evidentiary elements. This information is then used to infer the doubts and proof directions of the target ecological and environmental protection case, gradually constructing the case's evidentiary chain. The identified key evidentiary elements are then evaluated and controlled within the professional field using domain knowledge. Each element is then proven and judged in accordance with relevant legal provisions, professional standards, and other relevant criteria to provide reliable evidence and judgment. Finally, the organized evidentiary elements and proof bases are cross-correlated and verified to ensure the accuracy and coherence of the evidentiary chain and the formation of a logical basis for proof. This improves the quality, reliability, efficiency, and accuracy of the evidence, laying the foundation for a key victory or mediation in the case.

[0116] Compared with the prior art, the embodiments of the present invention obtain key evidence data by adaptively sampling the physical information data and digital information data related to the target ecological and environmental protection cases, thereby greatly improving the efficiency of data acquisition; the key evidence data is input into the attention distribution model for feature extraction to obtain the key evidence element feature vector; the key evidence element feature vector is input into the evidence element recognition model for recognition to obtain the evidence element recognition result of the target ecological and environmental protection cases, which can more accurately screen out the characteristics of the key evidence elements, improve the accuracy and sufficiency of the evidence, help resolve legal disputes over the ecological environment, and help achieve judicial justice in ecological and environmental protection cases.

[0117] An embodiment of the present invention also provides a computer storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, a method for identifying evidence elements in ecological and environmental protection cases is implemented.

[0118] If the integrated module is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device capable of carrying the computer program code to a construction device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electric carrier signal, a telecommunication signal and a software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0119] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, a method for identifying evidence elements in ecological and environmental protection cases is implemented.

[0120] The terminal device may be a computing device such as a desktop computer, a notebook computer, a PDA, a server, a server cluster, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0121] The processor may be a central processing unit (CPU), other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0122] In some embodiments, the memory may be an internal storage unit of the terminal device, such as a hard disk or memory of the terminal device. In other embodiments, the memory may also be an external storage device of the terminal device, such as a plug-in hard disk equipped on the terminal device, a smart memory card (SMC, Smart Media Card), a secure digital (SD, Secure Digital) card, a flash card, etc. Furthermore, the memory may include both an internal storage unit of the terminal device and an external storage device. The memory is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program. The memory may also be used to temporarily store data that has been output or is to be output.

[0123] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the embodiment of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0124] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the embodiments of the present invention. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0125] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for identifying evidence elements in ecological and environmental protection cases, characterized by: include: Step 1: Collect physical information data and digital information data related to the target ecological and environmental protection case. The physical information data includes environmental monitoring data related to the target ecological and environmental protection case acquired using monitoring equipment, and the digital information data includes text information data, image information data, and video information data related to the target ecological and environmental protection case. Step 2: Establishing an adaptive sampling model including a sensitivity analysis module based on a Markov Chain Monte Carlo algorithm, inputting the physical information data and the digital information data into the adaptive sampling model, analyzing the importance and sensitivity of the physical information data and the digital information data through the sensitivity analysis module, and adaptively sampling the physical information data and the digital information data based on the importance and sensitivity to obtain key evidence data; Step 3: Inputting the key evidence data into an attention distribution model with a self-supervised learning mechanism for feature extraction, obtaining local spatial features of the key evidence data, global spatial relationships and temporal relationships in the key evidence data, and integrating the spatial features of the key evidence data and the global spatial relationships and temporal relationships in the key evidence data to obtain a key evidence element feature vector; the local spatial features of the key evidence data include texture features, shape features, pollution features, and structural features of the ecological environment; The global spatial relationship and temporal relationship are the relationship and temporal relationship of the context information in the key evidence data; Step 4: Input the key evidence element feature vector into the evidence element recognition model for recognition, and obtain the evidence element recognition result of the target ecological and environmental protection case; The attention distribution model includes: an embedding module, a long short-term memory module, a convolutional neural network module and a fusion module; The input end of the embedding module is connected to the input end of the long short-term memory module, the output end of the long short-term memory module is connected to the input end of the convolutional neural network module, the output end of the convolutional neural network module and the output end of the long short-term memory module are both connected to the input end of the fusion module, and the output end of the fusion module is connected to the input end of the evidence element recognition model.

2. The method for identifying evidence elements in ecological and environmental protection cases according to claim 1 is characterized in that: Before step 2, the method further includes: Preprocessing the physical information data and the digital information data to obtain preprocessed physical information data and digital information data; The preprocessed physical information data and the preprocessed digital information data are input into the adaptive sampling model for adaptive sampling.

3. The method for identifying evidence elements in ecological and environmental protection cases according to claim 2 is characterized in that: The adaptive sampling model includes: The Markov chain Monte Carlo module, sensitivity analysis module and adaptive sampling module are connected in sequence; Inputting the preprocessed physical information data and the preprocessed digital information data into the Markov chain Monte Carlo algorithm for sampling to obtain physical information data and digital information data that meet a preset probability distribution; Inputting the physical information data and digital information data that meet the preset probability into the sensitivity analysis module, and analyzing the importance and sensitivity of the physical information data and the digital information data that meet the preset probability distribution through the sensitivity analysis module to obtain the importance and sensitivity of the physical information data and the digital information data and sampling weights; The importance and sensitivity of the physical information data and the digital information data and the sampling weight are input into the adaptive sampling module, and the sampling probability of the physical information data and the digital information data is adaptively adjusted by the adaptive sampling module to obtain key evidence data.

4. The method for identifying evidence elements in ecological and environmental protection cases according to claim 3 is characterized in that: The attention distribution model also includes: Convolution module and pooling module; The input end of the convolution module is connected to the output end of the long short-term memory module, the output end of the convolution layer is connected to the input end of the pooling module, and the output end of the pooling module is connected to the input end of the convolutional neural network module.

5. The method for identifying evidence elements in ecological and environmental protection cases according to claim 4 is characterized in that: The step 3 includes: Inputting the key evidence data into the embedding module of the attention distribution model for conversion to obtain a vector of the key evidence data; Extracting features from the vector in the long short-term memory module to obtain sequence data features; Inputting the sequence data features into the convolution module for first feature extraction to obtain local spatial features and generate feature maps; Inputting the feature map into the pooling module for a second feature extraction to obtain structural features; Inputting the local spatial features and the structural features into the convolutional neural network module for feature extraction to obtain global spatial relationships and temporal relationships in key evidence data; The local spatial features, the structural features and the global spatial relationships and temporal relationships in the key evidence data are input into the fusion module for weighted fusion to obtain a key evidence element feature vector.

6. The method for identifying evidence elements in ecological and environmental protection cases according to claim 5 is characterized in that: Before inputting the key evidence element feature vector into the evidence element recognition model for recognition, the method further includes: Classifying the key evidence element feature vectors by a classifier to obtain text feature vectors, image feature vectors, and video feature vectors; The text feature vector, the image feature vector and the video feature vector are input into an evidence element recognition model for recognition.

7. The method for identifying evidence elements in ecological and environmental protection cases according to claim 6 is characterized in that: The evidence element identification model includes: A first conversion module, a second conversion module, a third conversion module, a multi-task learning module, and an attention module; The input end of the first conversion module, the input end of the second conversion module, and the input end of the third conversion module are all connected to the output end of the classifier, the output end of the first conversion module, the output end of the second conversion module, and the output end of the third conversion module are all connected to the input end of the multi-task learning module, and the output end of the multi-task learning module is connected to the input end of the attention module; Inputting the text feature vector, the image feature vector and the video feature vector into the first conversion module, the second conversion module and the third conversion module respectively for conversion to obtain text data, image data and key frame images; Inputting the text data, the image data, and the key frame image into the multi-task learning module for multi-task learning to obtain learned text data, learned image data, and learned key frame image; The learned text data, learned image data and learned key frame images are input into the attention module for recognition to obtain the evidence element recognition results of the target ecological and environmental protection case. The evidence element recognition results include: the category prediction probability distribution of the learned text data, the learned image data and category labels, and the key frames of the learned key frame images.

8. The method for identifying evidence elements in ecological and environmental protection cases according to claim 7 is characterized in that: After step 4, the method further includes: A graph neural network is used to extract the relationship between the evidence elements in the evidence element recognition result to obtain the correlation relationship between the evidence elements; Classify and sort out the correlations between the various elements of evidence to obtain classification results; Based on the classification results, frequent patterns between the evidence elements are found through association rule learning algorithms to obtain key events; Based on the aforementioned correlations and key events, obtain the doubts and proof directions of the target ecological and environmental protection case; Based on the doubts and their proof direction, relevant information and clues are extracted to construct the chain of evidence for the target ecological and environmental protection case.

9. A computer storage medium, wherein the computer readable storage medium stores a computer program, characterized in that: When the computer program is executed by a processor, the method for identifying evidence elements in ecological and environmental protection cases as described in any one of claims 1 to 7 is implemented.

10. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for identifying evidence elements in ecological and environmental protection cases as described in any one of claims 1 to 7 is implemented.

Images