A neural network-based intelligent method and system for evaluating the credibility of evidence

Abstract

This invention relates to an intelligent method and system for assessing the credibility of evidence based on neural networks. The method includes: processing textual evidence to obtain processed text; extracting time data, location data, person data, and event data; constructing text time nodes, text location nodes, text person nodes, and text event nodes; constructing a set of text nodes; constructing a text relationship graph; setting a dynamic acquisition frequency; performing image extraction on video evidence to obtain an extracted image set; constructing video time nodes, video location nodes, video person nodes, video event nodes, a set of video graph edges, and a video relationship graph; setting the evidence state; calculating the evidence similarity; if the evidence similarity is less than a similarity threshold, the evidence state is false; if the evidence similarity is not less than the similarity threshold, the evidence state is credible. This invention can improve the efficiency of evidence credibility assessment.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to an intelligent method and system for evaluating the credibility of evidence based on neural networks. Background Technology

[0002] With the rapid development of artificial intelligence and data analysis technology, intelligent assessment of evidence credibility is widely used in fields such as justice, news investigation and social security. In modern case analysis, textual evidence and video evidence play a crucial role. Textual evidence records the key elements of an event through structured language, such as time, place, people and events, while video evidence provides real images of the scene, which helps to intuitively show the course of the case.

[0003] Traditional methods for assessing the credibility of evidence typically rely on human judgment and experience-based analysis, making them inefficient for handling massive amounts of evidence. For textual evidence, traditional methods primarily determine reliability through keyword matching and human reasoning, which not only depends on human bias but also struggles to handle complex and diverse evidence. In video evidence analysis, traditional methods largely depend on simple image processing techniques, such as edge detection and target tracking, failing to effectively extract and correlate crucial information from videos, resulting in low accuracy and efficiency in evidence analysis. Therefore, improving the efficiency of evidence credibility assessment is a pressing issue that needs to be addressed. Summary of the Invention

[0004] This invention provides a method and system for intelligent evaluation of evidence credibility based on neural networks, the main purpose of which is to improve the efficiency of evidence credibility evaluation.

[0005] To achieve the above objectives, this invention provides an intelligent evaluation method for evidence credibility based on neural networks, comprising:

[0006] Obtain textual and video evidence, perform text preprocessing based on the textual evidence to obtain processed text, perform structural analysis on the processed text, and extract data from the processed text after structural analysis based on a pre-built entity recognition model to obtain time data, location data, person data, and event data.

[0007] Based on the time data, location data, person data, and event data, construct text time nodes, text location nodes, text person nodes, and text event nodes;

[0008] A text node set is constructed based on the text time node, text location node, text character node, and text event node; and a text relationship graph is constructed based on the text node set.

[0009] A dynamic acquisition frequency is set, and the video evidence is image extracted based on the dynamic acquisition frequency and a pre-built convolutional neural network model to obtain an extracted image set. Based on the extracted image set, video time nodes, video location nodes, video character nodes, video event nodes and video graph edge sets are constructed, and a video relationship graph is constructed based on the video time nodes, video location nodes, video character nodes, video event nodes and video graph edge sets.

[0010] The evidence similarity is calculated based on the text relationship graph and the video relationship graph. The evidence status of the text evidence and the video evidence is judged based on the evidence similarity. The evidence status includes: credible status and false status.

[0011] Determine whether the similarity of the evidence is less than a preset similarity threshold;

[0012] If the similarity of the evidence is less than the similarity threshold, the evidence status is confirmed as false.

[0013] If the similarity of the evidence is not less than the similarity threshold, then the evidence status is confirmed as credible.

[0014] Optionally, constructing a text relationship graph based on a set of text nodes includes:

[0015] Based on the textual evidence, connection types are defined, including: time connection, location connection, person connection, and event connection. Time connection is used to connect text time nodes and text event nodes, location connection is used to connect text location nodes and text event nodes, person connection is used to connect text person nodes and text event nodes, and event connection is used to connect different text event nodes.

[0016] Construct a text graph edge set based on the aforementioned connection types, time connections, location connections, person connections, and event connections;

[0017] Construct a text relationship graph based on the aforementioned set of text nodes and the set of text graph edges:

[0018] Q = (W, E)

[0019] Where Q refers to the text relation graph, W refers to the set of text nodes, and E refers to the set of text graph edges.

[0020] Optionally, setting the dynamic acquisition frequency includes:

[0021] Set the data acquisition time interval, and calculate the initial data acquisition frequency based on the data acquisition time interval:

[0022]

[0023] Where R1 refers to the initial sampling frequency, and t refers to the sampling time interval;

[0024] Based on the initial acquisition frequency, images are extracted from the video evidence to obtain an initial image set, and image difference values ​​are calculated based on the initial image set.

[0025] Set the maximum and minimum difference values, and set the dynamic acquisition frequency based on the maximum difference value, image difference value, minimum difference value, and initial acquisition frequency:

[0026]

[0027] Where R2 refers to the dynamic acquisition frequency, T refers to the image difference value, T1 refers to the highest difference value, and T2 refers to the lowest difference value.

[0028] Optionally, calculating the image difference value based on the initial image set includes:

[0029] An initial image sequence is obtained based on the initial image set, wherein the initial image sequence includes multiple initial images. A first initial image is extracted sequentially from the initial image sequence, and the following operations are performed on the extracted first initial image:

[0030] Based on the extracted first initial image, a second initial image is identified in the initial image sequence, wherein the second initial image is an initial image that is adjacent to and lags behind the first initial image;

[0031] Pixel detection is performed on the first initial image and the second initial image respectively to obtain the first pixel matrix and the second pixel matrix;

[0032] Set the pixel x-coordinate, pixel y-coordinate, and decomposition standard deviation, where the value of the decomposition standard deviation is 1 or 2;

[0033] Construct a decomposition function based on the pixel x-coordinate, pixel y-coordinate, and decomposition standard deviation:

[0034]

[0035] Where A(x, y, θ) refers to the decomposition function when the pixel's horizontal coordinate is x, the pixel's vertical coordinate is y, and the decomposition standard deviation is θ, x refers to the pixel's horizontal coordinate, y refers to the pixel's vertical coordinate, θ refers to the decomposition standard deviation, exp refers to the exponential function, and π refers to pi.

[0036] Based on the decomposition function and the decomposition standard deviation, the first pixel matrix and the second pixel matrix are decomposed to obtain the first decomposed pixel set and the second decomposed pixel set, respectively.

[0037] A first pixel difference set is calculated based on the first decomposed pixel set, and a second pixel difference set is calculated based on the second decomposed pixel set, wherein the first pixel difference set consists of a first pixel difference and a second pixel difference;

[0038] The first set of pixel differences is summarized to obtain a first grayscale image, and the second set of pixel differences is summarized to obtain a second grayscale image.

[0039] The image difference value is calculated based on the first grayscale image and the second grayscale image.

[0040] Optionally, the step of decomposing the first pixel matrix and the second pixel matrix based on the decomposition function and the decomposition standard deviation to obtain the first decomposed pixel set and the second decomposed pixel set includes:

[0041] The decomposition standard deviation is set to 1 to obtain the first decomposition function. Based on the first decomposition function, the first pixel matrix and the second pixel matrix are convolved respectively to obtain the first decomposed pixel and the second decomposed pixel.

[0042] The decomposition standard deviation is set to 2 to obtain the second decomposition function. Based on the second decomposition function, the first pixel matrix and the second pixel matrix are convolved respectively to obtain the first convolved pixel and the second convolved pixel.

[0043] The first decomposed pixel set is obtained by summing the first decomposed pixel and the first convolutional pixel. The second decomposed pixel set is obtained by summing the second decomposed pixel and the second convolutional pixel.

[0044] Optionally, summarizing the first pixel difference set to obtain the first grayscale image includes:

[0045] Based on the first pixel difference set, extract the first pixel difference and the second pixel difference, obtain the maximum value function, and calculate the grayscale pixel based on the first pixel difference, the second pixel difference, the maximum value function, the pixel horizontal coordinate, and the pixel vertical coordinate.

[0046] S(x,y)=ξ(d1(x,y),d2(x,y))

[0047] Where S(x, y) refers to the grayscale pixel with x as the horizontal coordinate and y as the vertical coordinate, ξ refers to the maximum value function, d1(x, y) refers to the first pixel difference with x as the horizontal coordinate and y as the vertical coordinate, and d2(x, y) refers to the second pixel difference with x as the horizontal coordinate and y as the vertical coordinate.

[0048] The first pixel difference set is summarized based on the grayscale pixels to obtain the first grayscale image.

[0049] Optionally, calculating the image difference value based on the first grayscale image and the second grayscale image includes:

[0050] Obtain the total number of pixels, and calculate the image error value based on the pixel x-coordinate, pixel y-coordinate, grayscale pixels, and total number of pixels:

[0051]

[0052] Where F refers to the image error value, S2(x,y) refers to the grayscale pixel with x-coordinate and y-coordinate in the second grayscale image, and N refers to the total number of pixels.

[0053] Calculate the first mean, second mean, first variance, second variance, and comprehensive covariance based on the first grayscale image and the second grayscale image;

[0054] Set a first stable parameter and a second stable parameter, and calculate image similarity based on the first mean, the second mean, the first variance, the second variance, the comprehensive covariance, the first stable parameter, and the second stable parameter:

[0055]

[0056] Where F2 refers to image similarity, δ1 refers to the first mean, δ2 refers to the second mean, σ1 refers to the first stable parameter, ε1 refers to the first variance, ε2 refers to the second variance, σ2 refers to the second stable parameter, and ε3 refers to the comprehensive covariance.

[0057] Set the difference weight parameter, and calculate the image difference value based on the image error value, image similarity, and difference weight parameter:

[0058] F3=α×F+(1-α)×F2

[0059] Where F3 refers to the image difference value, and α refers to the difference weight parameter.

[0060] Optionally, calculating the evidence similarity based on the text relationship graph and the video relationship graph includes:

[0061] Calculate the similarity of time nodes, and also calculate the similarity of location nodes, person nodes, and event nodes;

[0062] The text graph edge set and the video graph edge set are transformed respectively to obtain text edge vectors and video edge vectors;

[0063] Set the edge vector index, and calculate the graph edge similarity based on the edge vector index, text edge vector, and video edge vector:

[0064]

[0065] Where k refers to the graph edge similarity, and i refers to the edge vector index. The text edge vector with edge vector index i, μ i The video side vector at index i, where ||*|| indicates the modulo operator. Point to the dot product symbol;

[0066] Set weights for time nodes, locations, people, events, and edge similarity. Calculate evidence similarity based on these weights, along with graph edge similarity.

[0067] P=α1×γ1+β2×γ2+β3×γ3+β4×γ4+β5×k

[0068] Where P refers to evidence similarity, =1 refers to time node weight, γ1 refers to time node similarity, β2 refers to location node weight, γ2 refers to location node similarity, β3 refers to person node weight, γ3 refers to person node similarity, β4 refers to event node weight, γ4 refers to event node similarity, and β5 refers to edge similarity weight.

[0069] Optionally, the calculation of time node similarity includes:

[0070] The text time nodes and video time nodes are summarized to obtain a set of text time nodes and a set of video time nodes.

[0071] Calculate the time node similarity based on the aforementioned text time node set and video time node set:

[0072]

[0073] Where H1 refers to the set of text time nodes, H2 refers to the set of video time nodes, ∩ refers to the intersection symbol, and ∪ refers to the union symbol.

[0074] To achieve the above objectives, the present invention also provides an intelligent evidence credibility assessment system based on neural networks, comprising:

[0075] The text relationship graph construction module is used to acquire text evidence and video evidence, perform text preprocessing based on the text evidence to obtain processed text, perform structural analysis on the processed text, and extract data from the structurally analyzed processed text based on a pre-built entity recognition model to obtain time data, location data, person data, and event data; construct text time nodes, text location nodes, text person nodes, and text event nodes based on the time data, location data, person data, and event data; construct a text node set based on the text time nodes, text location nodes, text person nodes, and text event nodes; and construct a text relationship graph based on the text node set.

[0076] The video relationship graph construction module is used to set a dynamic acquisition frequency, extract images from the video evidence based on the dynamic acquisition frequency and a pre-built convolutional neural network model to obtain an extracted image set, construct video time nodes, video location nodes, video character nodes, video event nodes and video graph edge set based on the extracted image set, and construct a video relationship graph based on the video time nodes, video location nodes, video character nodes, video event nodes and video graph edge set.

[0077] The evidence similarity calculation module is used to calculate the evidence similarity based on the text relationship graph and the video relationship graph, and to determine the evidence status of the text evidence and video evidence based on the evidence similarity. The evidence status includes: credible status and false status.

[0078] The evidence status judgment module is used to determine whether the evidence similarity is less than a preset similarity threshold. If the evidence similarity is less than the similarity threshold, the evidence status is confirmed as false. If the evidence similarity is not less than the similarity threshold, the evidence status is confirmed as credible.

[0079] To address the above problems, the present invention also provides an electronic device, the electronic device comprising:

[0080] Memory, storing at least one instruction;

[0081] The processor executes the instructions stored in the memory to implement the neural network-based intelligent evaluation method for evidence credibility described above.

[0082] To address the aforementioned problems, the present invention also provides a computer-readable storage medium storing at least one instruction, which is executed by a processor in an electronic device to implement the aforementioned neural network-based intelligent evaluation method for evidence credibility.

[0083] To address the problems described in the background art, this invention first preprocesses the textual evidence to obtain processed text. Then, it performs structural analysis on the processed text and extracts data from it based on a pre-built entity recognition model. The text preprocessing effectively removes irrelevant information and interference, including redundant spaces, punctuation marks, and stop words, and performs precise word segmentation. Removing redundant spaces, punctuation marks, and stop words effectively reduces noise in the processed text and the computational load of the entity recognition model. Precise word segmentation transforms the textual evidence into independent words or phrases, which is beneficial for the entity recognition model to focus more on specific content. Especially in multilingual or specialized texts, precise word segmentation helps identify important information without being disturbed by irrelevant symbols or words. Structural analysis allows for a better understanding of the hierarchical structure and grammatical relationships of the textual evidence, providing clear data for the entity recognition model and improving its accuracy and efficiency. Secondly, based on the extracted textual time nodes, A text node set is constructed by including text location nodes, text person nodes, and text event nodes. Based on this set, a text relationship graph is built. This graph clearly represents all data and relationships within the text evidence, effectively organizing and storing information, and providing a clear node structure for calculating evidence similarity. Next, a dynamic acquisition frequency is set, and image extraction is performed on the video evidence using this frequency and a convolutional neural network model. The dynamic acquisition frequency can be adjusted according to the content of the video evidence, improving acquisition efficiency, avoiding redundant image extraction, and saving computational resources. Then, a video relationship graph is constructed, providing precise relationship data for subsequent evidence similarity calculations, making the calculation more efficient and accurate. Finally, evidence similarity is calculated, and the evidence status is determined based on this similarity. Automated calculation of evidence similarity and similarity threshold judgment reduces human error and improves the accuracy and efficiency of evidence evaluation. Therefore, this invention can improve the efficiency of evidence credibility assessment. Attached Figure Description

[0084] Figure 1 This is a flowchart illustrating an embodiment of the intelligent evidence credibility assessment method based on neural networks provided by the present invention.

[0085] Figure 2 A functional block diagram of an intelligent evidence credibility assessment system based on neural networks provided in an embodiment of the present invention;

[0086] Figure 3 This is a schematic diagram of the structure of an electronic device that implements the neural network-based intelligent evaluation method for evidence credibility, as provided in an embodiment of the present invention.

[0087] Explanation of reference numerals in the attached figures:

[0088] 10. Electronic device; 11. Processor; 12. Memory; 13. Bus.

[0089] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0090] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0091] This application provides a neural network-based intelligent evaluation method for evidence credibility. The executing entity of this neural network-based intelligent evaluation method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the neural network-based intelligent evaluation method for evidence credibility can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0092] Example 1:

[0093] Reference Figure 1 As shown in the diagram, this embodiment provides a flowchart of an intelligent evidence credibility assessment method based on neural networks. The intelligent evidence credibility assessment method based on neural networks includes:

[0094] S1. Obtain text evidence and video evidence, perform text preprocessing based on the text evidence to obtain processed text, perform structural analysis on the processed text, and extract data from the processed text after structural analysis based on a pre-built entity recognition model to obtain time data, location data, person data, and event data.

[0095] Interpretable, text evidence refers to evidence presented in text form. For example, in a news report, it is mentioned that "On October 1, 2024, Company A located in Shanghai announced the launch of a new product production line." Video evidence refers to evidence presented in video form. For example, a video recorded by a bank's surveillance equipment. Text preprocessing of text evidence refers to performing word segmentation, stop word removal, and syntactic parsing on text evidence. Word segmentation means decomposing text evidence into independent words or phrases for easy analysis. Stop word removal means removing words that do not contribute to the semantics, such as "is", "of", etc. Syntactic parsing means identifying the grammatical structure of text evidence and parsing the subject, predicate, and object components. For example, the text evidence is: "On December 12, 2024, Zhang planned a meeting in London and discussed the project budget with Zhao." After performing text preprocessing on the text evidence, the processed text is obtained: "On December 12, 2024, Zhang, London, planned a meeting, Zhao, discussed the budget." The processed text refers to the text obtained after performing text preprocessing on text evidence. For example, the processed text: "On December 12, 2024, Zhang, London, planned a meeting, Zhao, discussed the budget." Structural analysis of the processed text means dividing the processed text according to four dimensions: time, place, person, and event. For example, the processed text is: "On December 12, 2024, Zhang, London, planned a meeting, Zhao, discussed the budget." After performing structural analysis on the processed text, the result is: Time: December 12, 2024, Place: London, Persons: Zhang, Zhao, Events: planned a meeting, discussed the budget. The entity recognition model refers to BERT - Bi - LSTM - CRF, which is an entity recognition model that combines BERT, Bi - LSTM, and CRF. The entity recognition model is used to perform data recognition on the processed text and extract time data, place data, person data, and event data. Time data refers to the data in the text evidence used to represent time, such as "December 12, 2024". Place data refers to the data in the text evidence used to represent place, such as "London". Person data refers to the data in the text evidence used to represent persons, such as "Zhang, Zhao". Event data refers to the data in the text evidence used to represent specific events, such as "planned a meeting, discussed the budget".

[0096] S2. Construct text time nodes, text place nodes, text person nodes, and text event nodes according to the time data, place data, person data, and event data.

[0097] Understandably, a text time node refers to a node constructed based on time data, used to store time data and represent the specific time in the text evidence; a text location node refers to a node constructed based on location data, used to store location data and represent the specific location in the text evidence; a text person node refers to a node constructed based on person data, used to store person data and represent the specific person in the text evidence; and a text event node refers to a node constructed based on event data, used to store event data and represent the specific event in the text evidence.

[0098] S3. Construct a text node set based on the text time node, text location node, text person node, and text event node, and construct a text relationship graph based on the text node set.

[0099] Explainable, a text node set refers to the collection of text time nodes, text location nodes, text person nodes, and text event nodes. A text relationship graph refers to a structured graph formed by connecting text time nodes, text location nodes, text person nodes, and text event nodes through relationship types. It aims to show the relationships between time data, location data, person data, and event data, providing support for further analysis or reasoning. Relationship type refers to the type of relationship between text time nodes, text location nodes, text person nodes, and text event nodes. Relationship types include: time relationship, location relationship, person relationship, and event relationship. Time relationship is used to connect text time nodes and text event nodes, location relationship is used to connect text location nodes and text event nodes, person relationship is used to connect text person nodes and text event nodes, and event relationship is used to connect different text event nodes.

[0100] Specifically, the construction of a text relationship graph based on a set of text nodes includes:

[0101] Based on the textual evidence, connection types are defined, including: time connection, location connection, person connection, and event connection. Time connection is used to connect text time nodes and text event nodes, location connection is used to connect text location nodes and text event nodes, person connection is used to connect text person nodes and text event nodes, and event connection is used to connect different text event nodes.

[0102] Construct a text graph edge set based on the aforementioned connection types, time connections, location connections, person connections, and event connections;

[0103] Construct a text relationship graph based on the aforementioned set of text nodes and the set of text graph edges:

[0104] Q = (W, E)

[0105] Where Q refers to the text relation graph, W refers to the set of text nodes, and E refers to the set of text graph edges.

[0106] Explainable, the text graph edge set refers to the set formed by the edges of the text graph. A text graph edge is an edge that connects two nodes in a text relationship graph. For example, if the person data in the text person node is John and the event data in the text event node is Planning Meeting, then (John, Planning Meeting) is the edge connecting the text person node and the text event node.

[0107] S4. Set a dynamic acquisition frequency, extract images from the video evidence based on the dynamic acquisition frequency and a pre-built convolutional neural network model to obtain an extracted image set, construct video time nodes, video location nodes, video character nodes, video event nodes and video graph edge set based on the extracted image set, and construct a video relationship graph based on the video time nodes, video location nodes, video character nodes, video event nodes and video graph edge set.

[0108] Explainable terms include: dynamic acquisition frequency (MPF) refers to the frequency of image extraction from video evidence, which is dynamically adjusted based on the magnitude of image differences; extracted image set refers to the set of images obtained after image extraction from video evidence; video time node refers to a node constructed based on time data in video evidence; video location node refers to a node constructed based on location data in video evidence; video person node refers to a node constructed based on person data in video evidence; video event node refers to a node constructed based on event data in video evidence; video graph edge set refers to the set formed by video graph edges; video graph edge refers to the edge connecting two nodes in the video relationship graph. For example, if the time data in the video time node is December 12, 2024, and the event data in the video event node is "planning a meeting," then December 12, 2024 ("planning a meeting") is the edge connecting the video time node and the video event node. The video relationship graph refers to a structured graph formed by connecting video time nodes, video location nodes, video person nodes, and video event nodes through the video graph edge set.

[0109] Specifically, setting the dynamic acquisition frequency includes:

[0110] Set the data acquisition time interval, and calculate the initial data acquisition frequency based on the data acquisition time interval:

[0111]

[0112] Where R1 refers to the initial sampling frequency, and t refers to the sampling time interval;

[0113] Based on the initial acquisition frequency, images are extracted from the video evidence to obtain an initial image set, and image difference values ​​are calculated based on the initial image set.

[0114] Set the maximum and minimum difference values, and set the dynamic acquisition frequency based on the maximum difference value, image difference value, minimum difference value, and initial acquisition frequency:

[0115]

[0116] Where R2 refers to the dynamic acquisition frequency, T refers to the image difference value, T1 refers to the highest difference value, and T2 refers to the lowest difference value.

[0117] Explainable, the acquisition time interval refers to the time interval between extracting two frames of images from a video. For example, if the acquisition time interval is set to 0.03s, then the time interval between two adjacent frames extracted from the video is 0.03s. The initial acquisition frequency refers to the frequency at which preliminary image extraction is performed on video evidence. The initial image set refers to the set of initial images. The initial image refers to the image obtained by extracting video evidence according to the initial acquisition frequency. The image difference value refers to the value that quantifies the difference between two consecutive images. The highest difference value refers to the maximum value of the difference between two images, and the lowest difference value refers to the minimum value of the difference between two images. When the image difference value is greater than the highest difference value, it indicates that there is a significant difference between the two images, and the image acquisition frequency needs to be increased. When the image difference value is less than the lowest difference value, it indicates that there is no significant difference between the two images, and the degree of similarity is high, and the image acquisition frequency needs to be reduced to minimize image duplication.

[0118] Specifically, the calculation of image difference values ​​based on the initial image set includes:

[0119] An initial image sequence is obtained based on the initial image set, wherein the initial image sequence includes multiple initial images. A first initial image is extracted sequentially from the initial image sequence, and the following operations are performed on the extracted first initial image:

[0120] Based on the extracted first initial image, a second initial image is identified in the initial image sequence, wherein the second initial image is an initial image that is adjacent to and lags behind the first initial image;

[0121] Pixel detection is performed on the first initial image and the second initial image respectively to obtain the first pixel matrix and the second pixel matrix;

[0122] Set the pixel x-coordinate, pixel y-coordinate, and decomposition standard deviation, where the value of the decomposition standard deviation is 1 or 2;

[0123] Construct a decomposition function based on the pixel x-coordinate, pixel y-coordinate, and decomposition standard deviation:

[0124]

[0125] Where A(x, y, θ) refers to the decomposition function when the pixel's horizontal coordinate is x, the pixel's vertical coordinate is y, and the decomposition standard deviation is θ, x refers to the pixel's horizontal coordinate, y refers to the pixel's vertical coordinate, θ refers to the decomposition standard deviation, exp refers to the exponential function, and π refers to pi.

[0126] Based on the decomposition function and the decomposition standard deviation, the first pixel matrix and the second pixel matrix are decomposed to obtain the first decomposed pixel set and the second decomposed pixel set, respectively.

[0127] A first pixel difference set is calculated based on the first decomposed pixel set, and a second pixel difference set is calculated based on the second decomposed pixel set, wherein the first pixel difference set consists of a first pixel difference and a second pixel difference;

[0128] The first set of pixel differences is summarized to obtain a first grayscale image, and the second set of pixel differences is summarized to obtain a second grayscale image.

[0129] The image difference value is calculated based on the first grayscale image and the second grayscale image.

[0130] Explainable, the initial image sequence refers to the sequence obtained by sorting the initial images in the initial image set according to their extraction time from earliest to latest. For example, 1, the first initial image, 2, the second initial image, where 1 and 2 constitute the initial image sequence. The first and second initial images refer to the images extracted sequentially from the initial image set, with the first initial image being extracted before the second. The first pixel matrix refers to the pixel matrix of the first initial image, and the second pixel matrix refers to the pixel matrix of the second initial image. A pixel matrix is ​​a two-dimensional array where each element represents a pixel in the image. The pixel x-coordinate refers to the x-coordinate of the pixel matrix, and the pixel y-coordinate refers to the y-coordinate of the pixel matrix. The decomposition standard deviation is a parameter that controls the image decomposition in the initial image set. The number of selectable values ​​for the decomposition standard deviation determines the number of layers for image decomposition. For example, if the decomposition standard deviation is 1 or 2, and the number is 2, the images in the initial image set are decomposed into two layers. When the decomposition standard deviation is 1, the initial image is decomposed using a decomposition function to obtain the first layer image. When the decomposition standard deviation is 2, the initial image is decomposed using the decomposition function to obtain the second layer image. The initial image refers to the image in the initial image set. The size of the first layer image is one-quarter of the size of the initial image, and the size of the second layer image is one-sixteenth of the size of the initial image. The first layer image refers to the image obtained after decomposing the initial image using the decomposition function when the decomposition standard deviation is 1. The second layer image refers to the image obtained after decomposing the initial image using the decomposition function when the decomposition standard deviation is 2. The decomposition function refers to the function that decomposes the initial image. The first decomposition pixel set refers to the set formed by the first decomposition pixel and the first convolution pixel. The first decomposition pixel refers to the pixel of the first layer image corresponding to the first pixel image. The first convolution pixel refers to the pixel of the second layer image corresponding to the first pixel image. The first pixel image refers to the image corresponding to the first pixel matrix. The second decomposition pixel set refers to the set formed by the second decomposition pixel and the second convolution pixel. The second decomposition pixel refers to the pixel of the first layer image corresponding to the second pixel image. The second convolution pixel refers to the pixel of the second layer image corresponding to the second pixel image. The second pixel image refers to the image corresponding to the second pixel matrix.

[0131] Understandably, the first pixel difference set refers to the set formed by the first pixel difference and the second pixel difference. The first pixel difference refers to the difference between the pixel in the first expanded pixel and the corresponding pixel in the first initial image. The first expanded pixel refers to the pixels of the image obtained after expanding the size of the image corresponding to the first decomposed pixel to the same size as the first initial image. Expansion means repeating the pixels in the first decomposed pixel, and repetition means copying the pixels in the first decomposed pixel and reusing the copied pixels to achieve the purpose of expanding the size. For example, the first decomposed pixel is: Further expansion yields: To expand refers to copying and reuse Expand the first decomposed pixel as 255, 254, 253, and 252 are all pixels. The second pixel difference refers to the difference between the pixels in the second expanded pixel and the corresponding pixels in the first decomposed pixel. The second expanded pixel refers to the pixels of the image obtained after expanding the size of the image corresponding to the first convolution pixel to the same size as the first decomposed pixel. The first initial image refers to the image corresponding to the first decomposed pixel before decomposition. The first grayscale image refers to the image obtained after integrating the first pixel difference set. The first grayscale image is a grayscale image of the same size as the first initial image. The second grayscale image refers to a grayscale image of the same size as the second initial image. A grayscale image refers to an image that only contains grayscale information. The grayscale value of each pixel in the grayscale image is an integer between 0 and 255. Grayscale information refers to the brightness of each pixel in the grayscale image. The higher the grayscale value, the stronger the brightness. The second initial image refers to the image corresponding to the second decomposed pixel before decomposition. For example, the size of the first initial image is a 4×4 matrix, as shown below: The first initial image is decomposed to obtain the first decomposed pixels. Decomposing the first initial image means decomposing the pixels in the upper left quarter of the matrix corresponding to the first initial image. Therefore, the first decomposed pixels are: The first decomposed pixel is decomposed to obtain the first convolutional pixel. Decomposing the first decomposed pixel refers to decomposing the top-left quarter of the first decomposed pixel. The second decomposed pixel is:

[254] . The first decomposed pixel is expanded into a 4×4 first expanded pixel with the same size as the first initial image: Calculate the difference between each pixel in the first initial image and each pixel in the corresponding first expanded image to obtain the first pixel difference: Expand the second decomposed pixel to the same size as the first decomposed pixel: The expansion method for the second decomposed pixel is the same as that for the first decomposed pixel, and will not be repeated here. The difference between each pixel in the first decomposed pixel and each pixel in the corresponding second expanded pixel is calculated to obtain the second pixel difference: The second pixel difference is augmented to obtain Then and The first grayscale image is obtained by integrating the results using the maximum value function:

[0132] In detail, the step of decomposing the first pixel matrix and the second pixel matrix based on the decomposition function and the decomposition standard deviation to obtain the first decomposed pixel set and the second decomposed pixel set includes:

[0133] The decomposition standard deviation is set to 1 to obtain the first decomposition function. Based on the first decomposition function, the first pixel matrix and the second pixel matrix are convolved respectively to obtain the first decomposed pixel and the second decomposed pixel.

[0134] The decomposition standard deviation is set to 2 to obtain the second decomposition function. Based on the second decomposition function, the first pixel matrix and the second pixel matrix are convolved respectively to obtain the first convolved pixel and the second convolved pixel.

[0135] The first decomposed pixel set is obtained by summing the first decomposed pixel and the first convolutional pixel. The second decomposed pixel set is obtained by summing the second decomposed pixel and the second convolutional pixel.

[0136] Explained, the first decomposition function refers to the decomposition function when the decomposition standard deviation is set to 1, and the second decomposition function refers to the decomposition function when the decomposition standard deviation is set to 2. When the decomposition standard deviation is set to 1, the size of the convolved image is one-quarter of the size of the unconvolved image. When the decomposition standard deviation is set to 2, the size of the convolved image is one-sixteenth of the size of the unconvolved image. Convolution refers to the process of applying a convolution kernel to the input image to decompose the input image. The convolution kernel refers to the decomposition function. The input image refers to the first pixel matrix, the second pixel matrix, the first pixel matrix, and the second pixel matrix.

[0137] In detail, the step of summarizing the first pixel difference set to obtain the first grayscale image includes:

[0138] Based on the first pixel difference set, extract the first pixel difference and the second pixel difference, obtain the maximum value function, and calculate the grayscale pixel based on the first pixel difference, the second pixel difference, the maximum value function, the pixel horizontal coordinate, and the pixel vertical coordinate.

[0139] S(x,y)=ξ(d1(x,y),d2(x,y))

[0140] Where S(x, y) refers to the grayscale pixel with x as the horizontal coordinate and y as the vertical coordinate, ξ refers to the maximum value function, d1(x, y) refers to the first pixel difference with x as the horizontal coordinate and y as the vertical coordinate, and d2(x, y) refers to the second pixel difference with x as the horizontal coordinate and y as the vertical coordinate.

[0141] The first pixel difference set is summarized based on the grayscale pixels to obtain the first grayscale image.

[0142] Explainable, the maximum value function refers to the function that takes the maximum value, for example, ξ(2,5)=5, and grayscale pixel refers to the grayscale value of the pixel in the first grayscale image.

[0143] Specifically, the step of calculating the image difference value based on the first grayscale image and the second grayscale image includes:

[0144] Obtain the total number of pixels, and calculate the image error value based on the pixel x-coordinate, pixel y-coordinate, grayscale pixels, and total number of pixels:

[0145]

[0146] Where F refers to the image error value, S2(x,y) refers to the grayscale pixel with x-coordinate and y-coordinate in the second grayscale image, and N refers to the total number of pixels.

[0147] Calculate the first mean, second mean, first variance, second variance, and comprehensive covariance based on the first grayscale image and the second grayscale image;

[0148] Set a first stable parameter and a second stable parameter, and calculate image similarity based on the first mean, the second mean, the first variance, the second variance, the comprehensive covariance, the first stable parameter, and the second stable parameter:

[0149]

[0150] Where F2 refers to image similarity, δ1 refers to the first mean, δ2 refers to the second mean, σ1 refers to the first stable parameter, ε1 refers to the first variance, ε2 refers to the second variance, σ2 refers to the second stable parameter, and ε3 refers to the comprehensive covariance.

[0151] Set the difference weight parameter, and calculate the image difference value based on the image error value, image similarity, and difference weight parameter:

[0152] F3=α×F+(1-α)×F2

[0153] Where F3 refers to the image difference value, and α refers to the difference weight parameter.

[0154] Explained, the total number of pixels refers to the number of pixels in the first and second grayscale images, and the total number of pixels in the first and second grayscale images is equal. The image error value refers to the mean of the squared errors of the pixel values ​​between the first and second grayscale images, reflecting the overall degree of difference between the first and second grayscale images. The larger the image error value, the greater the degree of difference. The squared error of the pixel value refers to the square of the difference between the pixel values ​​at corresponding pixel positions in the first and second grayscale images, where the corresponding pixel position refers to the position in the first and second grayscale images respectively. For example, if the position of the first grayscale image is (1,1), then the corresponding pixel position of the second grayscale image is also (1,1). Both the first and second stability parameters are constants used to avoid zero denominators and ensure computational stability. The first mean refers to the average of all grayscale values ​​in the first grayscale image, and the second mean refers to the average of all grayscale values ​​in the second grayscale image. The first variance refers to the degree of deviation between the grayscale values ​​of the first grayscale image and the first mean, reflecting the dispersion of the grayscale values. A larger first variance indicates a greater deviation between the grayscale values ​​of the first grayscale image and the first mean, and a greater degree of dispersion of the grayscale values. The higher the degree of similarity, the greater the deviation of the grayscale values ​​in the first and second grayscale images from the second mean. The second variance indicates the degree of deviation between the grayscale values ​​and the second mean, reflecting the dispersion of the grayscale values. A larger second variance indicates a greater deviation between the grayscale values ​​and the second mean, and a higher degree of dispersion. The covariance represents the similarity of grayscale value changes between the first and second grayscale images. Grayscale value changes refer to the changes in the grayscale values ​​of corresponding pixels in the first and second grayscale images. The similarity degree indicates the degree of similarity in grayscale value changes; a higher similarity degree indicates a greater similarity between the corresponding pixels in the first and second grayscale images. The more similar the changes in grayscale values, the more similar the first grayscale image and the second grayscale image are. Image similarity refers to the numerical value describing the degree of similarity between the first grayscale image and the second grayscale image. The higher the image similarity, the more similar the first grayscale image and the second grayscale image are. The difference weight parameter refers to the parameter that adjusts the degree of influence of the image error value on the image difference value. The larger the difference weight parameter, the greater the influence of the image error value on the image difference value. Among them, the first variance is the variance of the grayscale values ​​in the first grayscale image, the comprehensive covariance is the covariance of the grayscale values ​​in the first grayscale image and the grayscale values ​​in the second grayscale image, and the second variance is the variance of the grayscale values ​​in the second grayscale image.

[0155] S5. Calculate the evidence similarity based on the text relationship graph and the video relationship graph, and determine the evidence status of the text evidence and video evidence based on the evidence similarity. The evidence status includes: credible status and false status.

[0156] Explainable evidence status refers to the state in which textual and video evidence is judged to be credible. Evidence similarity refers to the degree of similarity between textual and video evidence. The greater the evidence similarity, the greater the similarity between textual and video evidence, and the more credible the content of textual and video evidence. Evidence status includes: credible status and false status. credible status indicates that textual and video evidence is credible, while false status indicates that textual and video evidence is not credible.

[0157] Specifically, the step of calculating evidence similarity based on the text relationship graph and the video relationship graph includes:

[0158] Calculate the similarity of time nodes, and also calculate the similarity of location nodes, person nodes, and event nodes;

[0159] The text graph edge set and the video graph edge set are transformed respectively to obtain text edge vectors and video edge vectors;

[0160] Set the edge vector index, and calculate the graph edge similarity based on the edge vector index, text edge vector, and video edge vector:

[0161]

[0162] Where k refers to the graph edge similarity, and i refers to the edge vector index. The text edge vector with edge vector index i, μ i The video side vector at index i, where ||*|| indicates the modulo operator. Point to the dot product symbol;

[0163] Set weights for time nodes, locations, people, events, and edge similarity. Calculate evidence similarity based on these weights, along with graph edge similarity.

[0164] P=β1×γ1+β2×γ2+β3×γ3+β4×γ4+β5×k

[0165] Where P refers to evidence similarity, β1 refers to time node weight, γ1 refers to time node similarity, α2 refers to location node weight, γ2 refers to location node similarity, β3 refers to person node weight, γ3 refers to person node similarity, β4 refers to event node weight, γ4 refers to event node similarity, and β5 refers to edge similarity weight.

[0166] Explainable similarity includes: time node similarity (the degree of similarity between text and video time nodes in the text and video relationship graphs, respectively), location node similarity (the degree of similarity between text and video time nodes, and event node similarity), and person node similarity (the degree of similarity between event nodes). This refers to the similarity between text event nodes and video event nodes in text and video relationship graphs. Higher similarity indicates a greater degree of similarity between text and video event nodes. Text edge vectors quantify and describe the type of connections between text time nodes, text location nodes, text character nodes, and text event nodes. Video edge vectors quantify and describe the type of connections between video time nodes, video location nodes, video character nodes, and video event nodes. Text and video edge vectors are obtained by transforming the text graph edge set and video graph edge set, respectively. For example, textual evidence states that a meeting was held on December 12, 2024. (20 On December 12, 2024, a meeting was held. The edges connecting text time nodes and text event nodes are defined as: [held, 1]. "held" indicates the connection between the text time node and the text event node, and "1" indicates the number of times this connection appears in the text evidence. The edge vector index refers to the parameters used to traverse the text edge vectors and video edge vectors. Graph edge similarity refers to the degree of similarity between the text edge vector and the video edge vector; a higher graph edge similarity indicates a higher degree of similarity between the text edge vector and the video edge vector. Time node weight refers to the parameter that adjusts the degree of influence of time node similarity on evidence similarity; a larger time node weight indicates a greater influence of time node similarity on evidence similarity. Location node weight refers to the parameter that adjusts... The parameters for determining the impact of location node similarity on evidence similarity are as follows: Location node weight (higher weight indicates greater impact); Person node weight (higher weight indicates greater impact); Event node weight (higher weight indicates greater impact); and Edge similarity weight (higher weight indicates greater impact).

[0167] Specifically, the calculation of time node similarity includes:

[0168] The text time nodes and video time nodes are summarized to obtain a set of text time nodes and a set of video time nodes.

[0169] Calculate the time node similarity based on the aforementioned text time node set and video time node set:

[0170]

[0171] Where H1 refers to the set of text time nodes, H2 refers to the set of video time nodes, ∩ refers to the intersection symbol, and ∪ refers to the union symbol.

[0172] Explainable, the text time node set refers to the set of time parameters stored in the text time nodes, and the video time node set refers to the set of time parameters stored in the video time nodes. Time parameters refer to time data, such as January 1, 2024, and January 2, 2024.

[0173] Understandably, the methods for calculating location node similarity, person node similarity, and event node similarity are the same as those for calculating time node similarity: obtain the text location node set and the video location node set, and calculate the location node similarity based on the text location node set and the video location node set.

[0174]

[0175] Wherein, γ2 refers to the location node similarity, M1 refers to the set of text location nodes, and M2 refers to the set of video location nodes;

[0176] Obtain the text-based and video-based character node sets, and calculate the character node similarity based on these sets:

[0177]

[0178] Wherein, γ3 refers to the similarity of character nodes, M1′ refers to the set of text character nodes, and M2′ refers to the set of video character nodes;

[0179] Obtain the sets of text event nodes and video event nodes, and calculate the similarity of event nodes based on the sets of text event nodes and video event nodes:

[0180]

[0181] Where γ4 refers to the event node similarity, M1" refers to the set of text event nodes, and M2" refers to the set of video event nodes.

[0182] Explainable, the text location node set refers to the set of location parameters stored in text location nodes, and the video location node set refers to the set of location parameters stored in video location nodes. Location parameters refer to location data, such as "London". The text person node set refers to the set of person parameters stored in text person nodes, and the video person node set refers to the set of person parameters stored in video person nodes. Person parameters refer to person data, including but not limited to: name, age. The text event node set refers to the set of event parameters stored in text event nodes, and the video event node set refers to the set of event parameters stored in video event nodes. Event parameters refer to event data, such as "planning a meeting".

[0183] S6. Determine whether the similarity of the evidence is less than the preset similarity threshold.

[0184] Interpretable, the similarity threshold refers to a threshold used to judge the similarity of evidence. When the evidence similarity is not less than the similarity threshold, it is considered credible; when the evidence similarity is less than the similarity threshold, it is considered unreliable.

[0185] If the similarity of the evidence is less than the similarity threshold, then execute S7 to confirm that the evidence status is false.

[0186] If the evidence similarity is not less than the similarity threshold, then execute S8 to confirm that the evidence status is credible.

[0187] Therefore, this embodiment first extracts data by analyzing the structure of the processed text to effectively remove irrelevant information and interference, such as removing redundant spaces, punctuation marks, and stop words, and performs accurate word segmentation. Removing redundant spaces, punctuation marks, and stop words can effectively reduce noise in the processed text and reduce the computational load of the entity recognition model.

[0188] Furthermore, by accurately segmenting words, textual evidence can be transformed into independent words or phrases, which is beneficial for entity recognition models to identify and can focus more on specific content. Especially in multilingual or professional texts, accurate word segmentation helps to identify important information without being disturbed by irrelevant symbols or words. Structural analysis can better understand the hierarchical structure and grammatical relationships of textual evidence, providing clear data for entity recognition models and improving the recognition accuracy and efficiency of entity recognition models.

[0189] Furthermore, a text relationship graph is constructed based on the set of text nodes. This graph clearly represents all data and relationships within the textual evidence, effectively organizing and storing information and providing a clear node structure for calculating evidence similarity. Subsequently, image extraction is performed on the video evidence based on dynamic acquisition frequency and a convolutional neural network model. This process adjusts the extraction based on the video evidence content, improving acquisition efficiency, avoiding redundant image extraction, and saving computational resources. The video relationship graph also provides precise relationship data for subsequent evidence similarity calculations, making the calculation more efficient and accurate. Finally, automated calculation of evidence similarity and similarity threshold judgment reduces human error and improves the accuracy and efficiency of evidence evaluation.

[0190] Example 2:

[0191] like Figure 2 As shown in the diagram, this embodiment provides a functional block diagram of an intelligent evidence credibility assessment system 100 based on neural networks. This system 100 can be installed in an electronic device. Depending on the functions implemented, the intelligent evidence credibility assessment system 100 may include a text relationship graph construction module 101, a video relationship graph construction module 102, an evidence similarity calculation module 103, and an evidence status judgment module 104. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0192] The text relationship graph construction module 101 is used to acquire text evidence and video evidence, perform text preprocessing based on the text evidence to obtain processed text, perform structural analysis on the processed text, and extract data from the structurally analyzed processed text based on a pre-built entity recognition model to obtain time data, location data, person data, and event data; construct text time nodes, text location nodes, text person nodes, and text event nodes based on the time data, location data, person data, and event data; construct a text node set based on the text time nodes, text location nodes, text person nodes, and text event nodes; and construct a text relationship graph based on the text node set.

[0193] The video relationship graph construction module 102 is used to set a dynamic acquisition frequency, extract images from the video evidence based on the dynamic acquisition frequency and a pre-built convolutional neural network model to obtain an extracted image set, construct video time nodes, video location nodes, video character nodes, video event nodes and video graph edge set based on the extracted image set, and construct a video relationship graph based on the video time nodes, video location nodes, video character nodes, video event nodes and video graph edge set.

[0194] The evidence similarity calculation module 103 is used to calculate the evidence similarity based on the text relationship graph and the video relationship graph, and to determine the evidence status of the text evidence and the video evidence based on the evidence similarity. The evidence status includes: credible status and false status.

[0195] The evidence status judgment module 104 is used to determine whether the evidence similarity is less than a preset similarity threshold; if the evidence similarity is less than the similarity threshold, the evidence status is confirmed to be false; if the evidence similarity is not less than the similarity threshold, the evidence status is confirmed to be credible.

[0196] In detail, the modules in the neural network-based intelligent evidence credibility assessment system 100 described in this embodiment of the invention employ the same methods as described above when in use. Figure 1 The method used is the same as the neural network-based intelligent evaluation method for evidence credibility described above, and it can produce the same technical effect, so it will not be repeated here.

[0197] like Figure 3 The diagram shown is a structural schematic of an electronic device that implements a neural network-based intelligent evaluation method for evidence credibility, according to an embodiment of the present invention.

[0198] The electronic device 1 may include a processor 10, a memory 11 and a bus 12, and may also include a computer program stored in the memory 11 and capable of running on the processor 10, such as the neural network-based intelligent evaluation method program for evidence credibility described in Embodiment 1.

[0199] The memory 11 includes at least one type of readable storage medium, including flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, disk, optical disk, etc.

[0200] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits packaged with the same or different functions. It executes various functions of the electronic device 1 and processes data by running or executing programs or modules stored in the memory 11 (such as the neural network-based intelligent evaluation method program for evidence credibility described in Embodiment 1) and calling data stored in the memory 11.

[0201] The bus 12 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc.

[0202] Figure 3 Only electronic devices with components are shown; it will be understood by those skilled in the art that... Figure 3 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0203] The program for intelligent evaluation of evidence credibility based on neural networks, stored in the memory 11 of the electronic device 1, is a combination of multiple instructions. When run in the processor 10, it can implement the intelligent evaluation method for evidence credibility based on neural networks as described in Embodiment 1.

[0204] Specifically, the processor 10's implementation method for the above instructions can be found in [reference needed]. Figures 1 to 3 The descriptions of the relevant steps in the corresponding embodiments are not repeated here.

[0205] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor of an electronic device, can implement the neural network-based intelligent evaluation method for evidence credibility described in Embodiment 1.

[0206] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0207] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A neural network-based intelligent evaluation method for evidence credibility, characterized in that, The method includes: Obtain textual and video evidence, perform text preprocessing based on the textual evidence to obtain processed text, perform structural analysis on the processed text, and extract data from the processed text after structural analysis based on a pre-built entity recognition model to obtain time data, location data, person data, and event data. Based on the time data, location data, person data, and event data, construct text time nodes, text location nodes, text person nodes, and text event nodes; A text node set is constructed based on the text time node, text location node, text character node, and text event node; and a text relationship graph is constructed based on the text node set. A dynamic acquisition frequency is set, and the video evidence is image extracted based on the dynamic acquisition frequency and a pre-built convolutional neural network model to obtain an extracted image set. Based on the extracted image set, video time nodes, video location nodes, video character nodes, video event nodes and video graph edge sets are constructed, and a video relationship graph is constructed based on the video time nodes, video location nodes, video character nodes, video event nodes and video graph edge sets. The setting of the dynamic acquisition frequency includes: Set the data acquisition time interval, and calculate the initial data acquisition frequency based on the data acquisition time interval: in, The initial sampling frequency. Refers to the time interval between data collections; Based on the initial acquisition frequency, images are extracted from the video evidence to obtain an initial image set, and image difference values ​​are calculated based on the initial image set. Set the maximum and minimum difference values, and set the dynamic acquisition frequency based on the maximum difference value, image difference value, minimum difference value, and initial acquisition frequency: in, Refers to the dynamic acquisition frequency. Image difference values The highest difference value. Refers to the lowest difference value; The evidence similarity is calculated based on the text relationship graph and the video relationship graph. The evidence status of the text evidence and the video evidence is judged based on the evidence similarity. The evidence status includes: credible status and false status. Determine whether the similarity of the evidence is less than a preset similarity threshold; If the similarity of the evidence is less than the similarity threshold, the evidence status is confirmed as false. If the similarity of the evidence is not less than the similarity threshold, then the evidence status is confirmed as credible.

2. The intelligent evaluation method for evidence credibility as described in claim 1, characterized in that, The construction of a text relationship graph based on a set of text nodes includes: Based on the textual evidence, connection types are defined, including: time connection, location connection, person connection, and event connection. Time connection is used to connect text time nodes and text event nodes, location connection is used to connect text location nodes and text event nodes, person connection is used to connect text person nodes and text event nodes, and event connection is used to connect different text event nodes. Construct a text graph edge set based on the aforementioned time connections, location connections, person connections, and event connections; Construct a text relationship graph based on the aforementioned set of text nodes and the set of text graph edges: in, Refers to a textual relationship graph. Refers to the set of text nodes. Refers to the edge set of the text graph.

3. The intelligent evaluation method for evidence credibility as described in claim 2, characterized in that, Calculating time-node similarity includes: The text time nodes and video time nodes are summarized to obtain a set of text time nodes and a set of video time nodes. Calculate the time node similarity based on the aforementioned text time node set and video time node set: in, Refers to the set of text time nodes. Refers to the set of video time points. Intersection symbol The union symbol.

4. A neural network-based intelligent evidence credibility assessment system, characterized in that, The system includes: The text relationship graph construction module is used to acquire text evidence and video evidence, perform text preprocessing based on the text evidence to obtain processed text, perform structural analysis on the processed text, and extract data from the structurally analyzed processed text based on a pre-built entity recognition model to obtain time data, location data, person data, and event data; construct text time nodes, text location nodes, text person nodes, and text event nodes based on the time data, location data, person data, and event data; construct a text node set based on the text time nodes, text location nodes, text person nodes, and text event nodes; and construct a text relationship graph based on the text node set. The video relationship graph construction module is used to set a dynamic acquisition frequency, extract images from the video evidence based on the dynamic acquisition frequency and a pre-built convolutional neural network model to obtain an extracted image set, construct video time nodes, video location nodes, video character nodes, video event nodes and video graph edge set based on the extracted image set, and construct a video relationship graph based on the video time nodes, video location nodes, video character nodes, video event nodes and video graph edge set. The setting of the dynamic acquisition frequency includes: Set the data acquisition time interval, and calculate the initial data acquisition frequency based on the data acquisition time interval: in, The initial sampling frequency. Refers to the time interval between data collections; Based on the initial acquisition frequency, images are extracted from the video evidence to obtain an initial image set, and image difference values ​​are calculated based on the initial image set. Set the maximum and minimum difference values, and set the dynamic acquisition frequency based on the maximum difference value, image difference value, minimum difference value, and initial acquisition frequency: in, Refers to the dynamic acquisition frequency. Image difference values The highest difference value. Refers to the lowest difference value; The evidence similarity calculation module is used to calculate the evidence similarity based on the text relationship graph and the video relationship graph, and to determine the evidence status of the text evidence and video evidence based on the evidence similarity. The evidence status includes: credible status and false status. The evidence status judgment module is used to determine whether the evidence similarity is less than a preset similarity threshold. If the evidence similarity is less than the similarity threshold, the evidence status is confirmed as false. If the evidence similarity is not less than the similarity threshold, the evidence status is confirmed as credible.

Images