Image data verification system based on artificial intelligence

Abstract

The invention, which relates to the technical field of image data processing, discloses an artificial intelligence-based image data verification system comprising a data acquisition module, an image analysis module, an image identification module, an image verification module and a verification decision module. According to the method, time sequence boundary constraint setting is carried out based on cultural relic time sequences, a cultural relic image data set is constructed, extraction and evolution feature analysis are carried out on image feature data of different edge time sequences, cultural relic image evolution feature coefficients are generated, and an original feature database and a dynamic feature database are constructed. The method comprises the steps of generating a cultural relic image restoration influence coefficient, constructing an image evolution recognition model, performing comparison verification according to preset restoration conformity, performing comparison with a preset threshold value, judging conformity between a real-time cultural relic image and existing restoration features in a database, and verifying evolution time sequence consistency of the real-time image in combination with cultural relic time sequence boundary constraints. And the verification accuracy and reliability are further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing, and in particular to an image material verification system based on artificial intelligence. BACKGROUND

[0002] Image verification refers to a process of verifying certain facts, information or identities by analyzing, comparing and confirming image materials (such as photos, videos, etc.). In the field of cultural heritage protection, image materials such as cultural relic photos play a crucial role. These images not only record the state of cultural relics at different times, but also are key materials for studying the history and evolution of cultural relics. In cultural relic restoration, comparing image materials before and after restoration can visually observe the changes in the shape, color and texture of cultural relics. With the aid of image technology, problems that may exist in the restoration process can be found, such as whether the filling of restoration materials is smooth and uniform, whether the restoration part is naturally connected with the original cultural relic, and whether the original structure and material of the cultural relic have been damaged, etc., so that timely measures can be taken for adjustment and improvement. Currently, the traditional method for verifying cultural relics mainly relies on manual comparison or simple pixel-based algorithms. By observing with the naked eye or calculating image difference values, version consistency is determined. In the verification process, the focus is on the dominant features of restoration traces. However, the cross-period features of cultural relic images are influenced by various factors such as preservation environment and restoration technology. Traditional methods are difficult to distinguish between natural aging and human tampering-induced feature changes, lack the ability to identify hidden tampering behavior, resulting in high false and missed detection rates. Therefore, it is difficult to fully exploit the potential value of the data, and the effect of constructing a tampering detection feature library is not good. In view of the above technical defects, a solution is proposed. SUMMARY

[0003] The purpose of the present application is to extract and analyze the evolution characteristics of image feature data of different edge time series, discover the feature evolution law of cultural relics at different stages, construct an image evolution recognition model, compare and verify according to the preset restoration compliance degree, and judge the consistency of real-time cultural relic images with the existing restoration features in the database, thereby further improving the accuracy and reliability of the verification.

[0004] To achieve the above purpose, the present application adopts the following technical scheme: an image material verification system based on artificial intelligence, comprising a data acquisition module, an image analysis module, an image recognition module, an image verification module, and a verification decision module. The data acquisition module collects cultural relic image data in advance through a collection device, sets time sequence boundary constraints based on cultural relic time series, classifies image data, and constructs a cultural relic image data set. An image analysis module is configured to extract image feature data at different time sequences from the cultural relic image data set, obtain an image feature data set, generate cultural relic image evolution feature coefficients through evolution feature analysis, construct an original feature database and a dynamic feature database based on original features and restoration features, perform cultural relic feature restoration degree analysis, and generate cultural relic image restoration influence coefficients; An image recognition module is configured to obtain the image feature data set and the cultural relic image evolution feature coefficients, construct an image evolution recognition model in combination with the cultural relic image restoration influence coefficients, and output real-time cultural relic image evolution tracking results; An image verification module is configured to compare the real-time cultural relic image evolution tracking results with the restoration features, verify restoration compliance and comparison deviation values, generate comparison analysis results, verify time sequence consistency according to cultural relic time sequence boundary constraints, and generate an overall compliance verification report.

[0005] Further, a verification decision module is further included, which obtains the overall compliance verification report and performs credibility risk assessment according to the compliance, and specifically includes the following: S100, obtaining the overall compliance verification report and performing cultural relic overall compliance analysis, and performing analysis based on a preset verification rule; S101, the preset verification rule includes three verification mechanisms, when the overall compliance is greater than 0.9, the intelligent verification is successful, when the overall compliance is less than 0.7, the automatic verification is abnormal and the third-party verification is performed, and when the overall compliance is between 0.7 and 0.9, the automatic review is triggered to perform re-review verification; S102, obtaining the execution times of the three verification mechanisms, visualizing the execution times, integrating the execution times of the third-party verification and the re-review verification, and when the execution times are greater than a set execution time threshold, adjusting the overall compliance threshold.

[0006] Further, the cultural relic image data is divided and classified through time sequence boundary constraints, and the specific process is as follows: S200, obtaining historical archive data of the cultural relic, extracting key time nodes, mapping discrete time nodes to a unified time axis, labeling time accuracy, and constructing a cultural relic time standard axis according to the cultural relic time sequence; S201, setting the time sequence boundary constraint by taking the era transition point as the center to demarcate the front and rear time periods, and the edge time period is divided into a cultural relic stable period, a transition period and a mixed period, the cultural relic stable period is an era feature cultural relic without restoration and damage, the transition period is an era with mixed features and a feature cultural relic, and the mixed period is a cultural relic with features of the pre-restoration era and superimposed original cultural relic features and restoration features; S202, according to the time sequence boundary constraint condition, the cultural relic image data is classified dynamically, the cultural relic state stable period image set, the cross period image set and the mixed period image set are obtained, each group of images is arranged in ascending order of time, the coding structure is set according to the difference of the image set, the time identification and the image time arrangement sequence, the image data coding is sorted, the image data is stored in the database according to the coding rule, and the coding of each image data is mapped to the time standard axis, and the belonging time interval is marked.

[0007] Further, the cultural relic image evolution characteristic coefficient is generated, and the specific process is as follows: S300, the cultural relic image feature data set is obtained, the low-quality image is removed, the resolution and color space are unified, and the picture is light compensated according to the time sequence; S301, setting the cultural relic extraction feature matching point, extracting the image feature data of different edge time sequence, obtaining the image feature data set; S302, according to the image feature data set, the texture and color evolution characteristic analysis is carried out, the cultural relic image evolution characteristic coefficient is generated, the time sequence model of the feature is constructed, the evolution characteristic coefficient is calculated, and the calculation process is as follows: ; In the above formula, is the feature value of the current time point T, F (T-1) is the feature value history state of the last time point T-1, is the difference between the current time T and the last time T-1.

[0008] Further, according to the original unrepaired and repaired real-time feature recognition condition, the image feature data set is classified, and the specific process is as follows: S400, the image feature data set is obtained, the original and repaired state feature data is extracted according to the original unrepaired and repaired real-time feature recognition condition, and the original feature database and dynamic feature database are constructed according to the cultural relic defect type, repair time and process; S401, the repair area feature and the original defect feature are overlapped in space, the similarity is calculated according to the feature, the cultural relic feature repair degree is analyzed, and the comprehensive repair degree of the cultural relic is obtained; S402, the comprehensive repair degree of each time point in the repair process is fitted and analyzed to obtain the image repair influence coefficient of the cultural relic repair, and the calculation process is as follows: ; In the above formula, X is the image repair influence coefficient, is the color difference between the repaired cultural relic and the original cultural relic, is the set standard feature difference, To repair the surface feature difference between the post-repair cultural relics and the original cultural relics, For the preset surface feature difference value, W1 and W2 are the set weight coefficients, I={1, 2, 3, …, N}, N is the total number of feature points selected on the cultural relics for evaluation, F Y For the feature of the repaired cultural relics, F S For the feature of the original cultural relics.

[0009] Further, an image evolution identification model is constructed, and the specific process is as follows: S500, obtain the image feature data set and the cultural relic image evolution feature coefficient, combine the cultural relic image restoration influence coefficient, to obtain the cultural relic original image and the image after repair, as the sample input data of the image evolution identification model, and according to the cultural relic image evolution feature coefficient and the cultural relic image restoration influence coefficient, as the physical constraint condition of the image evolution identification model; S501, construct a neural network architecture based on the physical constraint condition, and construct a loss function of the initial network model, embed the equation residual gradient into the loss function, and constrain the output of the image evolution identification model; S502, divide the image feature data set into training set and test set according to the proportion of 8:2, adjust the number of hidden layer nodes according to the sample amount of training set, modify the last full connection layer, and the output dimension corresponds to the output dynamic tracking result.

[0010] Further, it also includes a feature analysis unit, which is used to identify the feature material of cultural relics and classify and define the cultural relic material, including the following contents; S600, extract cultural relic image features according to cultural relic material characteristics, assign initial weight values to cultural relic features of each material based on evolution analysis and the influence degree of restoration degree link; S601, establish a weight adjustment feedback mechanism, combine the historical restoration records of cultural relics, and regularly evaluate and adjust the weight parameters.

[0011] Further, the evolution time sequence consistency of real-time image is verified, and an overall compliance verification report is generated, and the specific process is as follows: S700, obtain the real-time cultural relic image evolution tracking result, compare it with the restoration features in the dynamic feature database after dynamic restoration feature extraction, obtain the compliance degree of the current dynamic restoration feature, and compare and verify according to the preset restoration compliance degree; S701, obtain the deviation value between the real-time image feature and the restoration feature in the dynamic feature database, and compare it with the preset threshold value, if the deviation value is greater than the preset threshold value, it is determined that the restoration process is abnormal, to generate a comparison analysis result; S702. Combining the temporal boundary constraints of cultural relics, conduct time matching degree analysis on the restoration features, verify the consistency of the evolution time of real-time images, and generate an overall conformity verification report.

[0012] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This AI-based image data verification system collects cultural relic image data through acquisition devices, sets temporal boundary constraints based on the time series of cultural relics, classifies and categorizes the image data to construct a cultural relic image dataset, extracts and analyzes the evolution features of image data from different edge time series, generates cultural relic image evolution feature coefficients, discovers the feature evolution patterns of cultural relics at different stages, classifies the image feature dataset based on original unrestored and restored conditions, constructs an original feature database and a dynamic feature database, generates a cultural relic image restoration influence coefficient by analyzing the degree of restoration of cultural relic features, and constructs an image evolution recognition model based on the cultural relic image restoration influence coefficient, compares and verifies according to the preset restoration conformity, detects the deviation value between real-time image features and restored features in the dynamic feature database, compares it with the preset threshold to determine the conformity of real-time cultural relic images with existing restored features in the database, and verifies the temporal consistency of real-time image evolution by combining cultural relic temporal boundary constraints, generating an overall conformity verification report, further improving the accuracy and reliability of verification. Attached Figure Description

[0013] Figure 1 A schematic diagram of the overall system structure of the present invention is shown. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Example 1: like Figure 1 As shown, the image data verification system based on artificial intelligence includes a data acquisition module, an image analysis module, an image recognition module, an image verification module, and a verification decision module. The data acquisition module collects cultural relic image data in advance through acquisition equipment, sets time-series boundary constraints based on the time series of cultural relics, classifies the image data, and constructs a cultural relic image dataset. An image analysis module is configured to extract image feature data at different time sequences from the cultural relic image data set, obtain an image feature data set, generate cultural relic image evolution feature coefficients through evolution feature analysis, construct an original feature database and a dynamic feature database based on original features and restoration features, perform cultural relic feature restoration degree analysis, and generate cultural relic image restoration influence coefficients; An image recognition module is configured to obtain the image feature data set and the cultural relic image evolution feature coefficients, construct an image evolution recognition model in combination with the cultural relic image restoration influence coefficients, and output real-time cultural relic image evolution tracking results; An image verification module is configured to compare the real-time cultural relic image evolution tracking results with the restoration features, verify restoration compliance and comparison deviation values, generate comparison analysis results, verify time sequence consistency according to cultural relic time sequence boundary constraints, and generate an overall compliance verification report.

[0016] The verification decision module is further configured to obtain the overall compliance verification report and perform credibility risk assessment according to the compliance, and specifically includes the following: S100, obtaining the overall compliance verification report and performing cultural relic overall compliance analysis, and performing analysis based on preset verification rules; S101, the preset verification rules include three verification mechanisms, when the overall compliance is greater than 0.9, the intelligent verification is successful, when the overall compliance is less than 0.7, the automatic verification is abnormal and the third-party verification is performed, and when the overall compliance is between 0.7 and 0.9, the automatic review is triggered to perform re-review verification; S102, obtaining the execution times of the three verification mechanisms, visualizing the execution times, integrating the execution times of the third-party verification and the re-review verification, and adjusting the overall compliance threshold when the execution times are greater than a set execution time threshold.

[0017] The cultural relic image data is divided and classified through time sequence boundary constraints, and the specific process is as follows: S200, obtaining historical archive data of cultural relics, extracting key time nodes, mapping discrete time nodes to a unified time axis, labeling time accuracy, and constructing a cultural relic time standard axis according to a cultural relic time sequence; S201, setting the time sequence boundary constraint by demarcating the front and rear time periods with the era transition point as the center, and dividing the edge time period into a cultural relic state stable period, a transition period and a mixed period, the cultural relic state stable period is an era feature cultural relic without restoration and damage, the transition period is an era with mixed features and a feature cultural relic, and the mixed period is a cultural relic with features of the pre-restoration era and superimposed original cultural relic features and restoration features; S202, according to the timing boundary constraint condition, the cultural relic image data is dynamically classified to obtain the cultural relic state stable period image set, the cross period image set and the mixed period image set, each group of images is arranged in ascending order of time, the coding structure is set according to the difference of the image set, the time identification and the image time arrangement sequence, the image data coding is sorted, the image data is stored in the database according to the coding rule, and the coding of each image data is mapped to the time standard axis, and the belonging time interval is marked.

[0018] The cultural relic image evolution characteristic coefficient is generated, and the specific process is as follows: S300, the cultural relic image feature data set is obtained, the low-quality image is removed, the resolution and color space are unified, and the pictures are light compensated according to the time sequence; S301, setting the cultural relic extraction feature matching point, extracting the image feature data of different edge time sequence, obtaining the image feature data set; S302, according to the image feature data set, the texture and color evolution characteristic analysis is carried out, the cultural relic image evolution characteristic coefficient is generated, the time sequence model of the feature is constructed, and the evolution characteristic coefficient is calculated, and the calculation process is as follows: ; In the above formula, is the feature value of the current time point T, F (T-1) is the feature value history state of the last time point T-1, is the difference between the current time T and the last time T-1.

[0019] According to the original unrepaired and repaired real-time feature recognition condition, the image feature data set is classified, and the specific process is as follows: S400, the image feature data set is obtained, the original and repaired state feature data is extracted according to the original unrepaired and repaired real-time feature recognition condition, and the original feature database and dynamic feature database are constructed according to the cultural relic defect type, repair time and process; S401, the repair area feature and the original defect feature are overlapped in space, the similarity is calculated according to the feature, the cultural relic feature repair degree is analyzed, and the comprehensive repair degree of the cultural relic is obtained; S402, the comprehensive repair degree of each time point in the repair process is fitted and analyzed to obtain the image repair influence coefficient of the cultural relic repair, and the calculation process is as follows: ; In the above formula, X is the image repair influence coefficient, is the color difference between the repaired cultural relic and the original cultural relic, is the set standard feature difference, To address the differences in surface features between the restored artifact and the original artifact, The preset surface feature difference is defined by W1 and W2, which are weighting coefficients. I = {1, 2, 3, ..., N}, where N is the total number of feature points selected on the artifact for evaluation. F Y To determine the characteristics of the restored artifact, F S These are characteristics of primitive artifacts.

[0020] The specific process of constructing an image evolution recognition model is as follows: S500: Obtain the image feature dataset and the cultural relic image evolution feature coefficients, and combine them with the cultural relic image restoration influence coefficients to obtain the original image and the restored image of the cultural relic, which serve as the sample input data for the image evolution recognition model. The cultural relic image evolution feature coefficients and the cultural relic image restoration influence coefficients serve as the physical constraints of the image evolution recognition model. S501. Construct a neural network architecture based on physical constraints, and construct the loss function of the initial network model. Embed the equation residual gradient into the loss function to constrain the output of the image evolution recognition model. S502. Divide the image feature dataset into a training set and a test set in an 8:2 ratio. Adjust the number of hidden layer nodes according to the sample size of the training set, modify the last fully connected layer, and output the dynamic tracking results corresponding to the output dimension.

[0021] In this scheme, the image evolution recognition model is constructed by the following hierarchy and connection relationships: Feature extraction sub-network: ResNet50 pre-trained on ImageNet is used as the backbone network, its original classification layer is removed and retained to the global average pooling layer; the input is a single cultural relic image (512×512×3), and the output is a 2048-dimensional feature vector; Temporal modeling sub-network: Following the feature extraction sub-network, it adopts a two-layer bidirectional LSTM network; each time step inputs a 2048-dimensional feature vector of the current image and outputs a 256-dimensional temporal feature; this sub-network is used to capture the evolution pattern of cultural relic features in the time dimension; Physical constraint embedding layer: The evolution feature coefficients of cultural relic images and the impact coefficients of cultural relic image restoration are used as external physical constraints. They are mapped through a fully connected layer with an input dimension of 2 and an output dimension of 128 to obtain the constraint feature vector. Feature fusion and output layer: The 256-dimensional temporal features output by the LSTM are concatenated with the 128-dimensional constraint feature vector to obtain a 384-dimensional fused feature. Then, it passes through two fully connected layers with 128 and 64 dimensions respectively, and finally through a linear output layer with an output dimension of 1 to obtain the real-time cultural relic image evolution tracking result. The real-time cultural relic image evolution tracking result is a continuous value, which represents the degree of matching between the current image sequence and the expected evolution pattern. Training steps and parameter settings: Parameter initialization The pre-trained ResNet50 weights are loaded for the feature extraction subnetwork, and the remaining network layers are initialized using Xavier. Training environment and optimization settings The optimizer Adam is used with an initial learning rate of 0.001, a step decay strategy that decays by a factor of 0.5 every 20 epochs, a batch size of 16, and a total training period of 100. If the validation set loss does not decrease for 10 consecutive periods, the training is terminated early. Loss function design: The total loss is composed of the mean square error loss and the physical constraint residual term:

[0022] In the above formula, is the true label, is the model prediction value, is the evolution feature coefficient constraint term, is the repair influence coefficient constraint term, is the constraint weight, set to 0.1. During training, small batches of samples are randomly and repeatedly extracted from the training set for training. After all training samples are extracted, it is a training period. After iterating for several periods, the image evolution recognition model is obtained. The trained image evolution recognition model is deployed in the image verification system. In actual verification, the real-time collection of cultural relic image sequences is input into the model, and the real-time cultural relic image evolution tracking result is output, The real-time cultural relic image evolution tracking result is a comprehensive quantitative evaluation value output by the image evolution recognition model after analyzing the input real-time cultural relic image sequence, including: Repair state compliance: the similarity between the current image features and the standard repair features recorded in the dynamic feature database; Temporal logic consistency: whether the current image state, such as the degree of damage and repair traces, conforms to the historical evolution stage it should have on the time sequence boundary constraints of the cultural relic time axis.

[0023] It also includes a feature analysis unit, which is used to identify the features of cultural relics and classify and define the materials of cultural relics, including the following content. S600, according to the features of cultural relic materials, extract the features of cultural relic images, based on evolution analysis and the influence of repair degree, assign initial weight values to the features of cultural relics of each material; S601, establish a weight adjustment feedback mechanism, combine the historical repair records of cultural relics, and regularly evaluate and adjust the weight parameters.

[0024] The evolution time sequence consistency of the real-time image is verified, and an overall compliance verification report is generated. The specific process is as follows: S700, the real-time cultural relic image evolution tracking result is obtained, and after dynamic repair feature extraction, the repair features in the dynamic feature database are compared to obtain the compliance of the current dynamic repair features. The repair compliance is compared according to the preset repair compliance; S701, the deviation value between the real-time image features and the repair features in the dynamic feature database is obtained, and compared with the preset threshold value. If the deviation value is greater than the preset threshold value, it is determined that the repair process is abnormal, and a comparison analysis result is generated; S702, combined with the time sequence boundary constraint of cultural relics, the time matching degree of the repair features is analyzed, the evolution time sequence consistency of the real-time image is verified, and an overall compliance verification report is generated.

[0025] The size of the interval and the threshold is set for easy comparison. The size of the threshold depends on the number of sample data and the base number set by the person skilled in the art for each group of sample data; as long as the proportion relationship between the image parameters and the quantized values is not affected.

[0026] The above formulas are dimensionless values calculated. The formula is obtained by software simulation of a large amount of data to obtain the most real situation. The preset parameters in the formula are set by the person skilled in the art according to the actual situation; In the two embodiments provided in the present application, it should be understood that the disclosed device and system can be implemented in other ways; for example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed; in addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be through some interface, indirect coupling or communication connection between the devices or modules, which can be electrical, mechanical or other forms; The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can make equivalent replacements or changes to the technical solutions and inventive concepts of the present application within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. An image data verification system based on artificial intelligence, characterized in that, It includes a data acquisition module, an image analysis module, an image recognition module, an image verification module, and a verification decision module; The data acquisition module collects cultural relic image data in advance through acquisition equipment, sets time-series boundary constraints based on the time series of cultural relics, classifies the image data, and constructs a cultural relic image dataset. The image analysis module is used to extract image feature data from cultural relic image datasets at different time series to obtain image feature datasets. Through evolution feature analysis, it generates cultural relic image evolution feature coefficients. Based on the original features and restoration features, it constructs an original feature database and a dynamic feature database, performs cultural relic feature restoration degree analysis, and generates cultural relic image restoration impact coefficients. The image recognition module acquires image feature datasets and cultural relic image evolution feature coefficients, combines them with cultural relic image restoration impact coefficients, and constructs an image evolution recognition model to output real-time cultural relic image evolution tracking results. The image verification module compares the real-time cultural relic image evolution tracking results with the restoration features to verify the restoration conformity and comparison deviation value, generates comparison analysis results, verifies the temporal consistency based on the cultural relic temporal boundary constraints, and generates an overall conformity verification report.

2. The image data verification system based on artificial intelligence according to claim 1, characterized in that, It also includes a verification decision module, which obtains an overall compliance verification report and performs a credibility risk assessment based on the compliance, specifically including the following: S100. Obtain the overall compliance verification report and conduct an overall compliance analysis of the cultural relics based on the preset verification rules. S101. The preset verification rules include three verification mechanisms. When the overall compliance is greater than 0.9, it means that the intelligent verification is successful. When the overall compliance is less than 0.7, it means that the automatic verification is abnormal and third-party verification is required. When the overall compliance is between 0.7 and 0.9, it means that automatic review is triggered for re-verification. S102. Obtain the execution counts of the three verification mechanisms, visualize the execution counts, and integrate the execution counts of third-party verification and re-verification. If the execution counts exceed the set threshold, adjust the overall compliance threshold.

3. The image data verification system based on artificial intelligence according to claim 1, characterized in that, The cultural relic image data is divided and classified using temporal boundary constraints. The specific process is as follows: S200. Obtain historical archive data of cultural relics, extract key time nodes, map discrete time nodes to a unified time axis, mark time precision, and construct a standard time axis for cultural relics based on the time sequence of cultural relics. S201. Delineate the time periods before and after the transition point of the era as the center, and set temporal boundary constraints. The edge time period is divided into the stable period of cultural relics, the cross period and the mixed period. The stable period of cultural relics is cultural relics with the characteristics of the era without restoration and without damage. The cross period is cultural relics with the characteristics of the era with mixed production techniques. The mixed period is cultural relics with the characteristics of the era before restoration, as well as cultural relics with the superimposed characteristics of the original cultural relics and the restoration characteristics. S202. Based on the temporal boundary constraints, the image data of cultural relics are dynamically classified to obtain image sets of the stable period, the cross-period, and the mixed period. Each group of images is arranged in ascending order of time. The encoding structure is set according to the different image sets, the era identifier, and the time sequence of the images. The image data is encoded and sorted. The image data is stored in the database according to the encoding rules. The encoding of each image data is mapped to the time standard axis and the time interval to which it belongs is marked.

4. The image data verification system based on artificial intelligence according to claim 1, characterized in that, The specific process for generating the evolution characteristic coefficients of cultural relic images is as follows: S300. Obtain the image feature dataset of cultural relics, remove low-quality images and unify the resolution and color space, and perform illumination compensation on the images according to the acquisition time series. S301. Set up matching points for cultural relic extraction features, extract image feature data from different edge time series, and obtain an image feature dataset; S302. Based on the image feature dataset, perform texture and color evolution feature analysis to generate cultural relic image evolution feature coefficients. Construct a time series model for the features to calculate the evolution feature coefficients. The calculation process is as follows: ； In the above formula, F(T) represents the eigenvalue at the current time point T, and F(T-1) represents the historical state of the eigenvalue at the previous time point T-1. It is the difference between the current time T and the previous time T-1.

5. The image data verification system based on artificial intelligence according to claim 1, characterized in that, Based on the criteria of original unrepaired and real-time repaired images, the image feature dataset is classified. The specific process is as follows: S400. Obtain the image feature dataset, extract feature data in the original and restored states based on the original unrepaired and real-time restored states as feature recognition conditions, and construct the original feature database and dynamic feature database according to the type of cultural relic defects, restoration time and process. S401. Spatially overlap the features of the repaired area with the features of the original defects, calculate the similarity based on the features, analyze the degree of restoration of the cultural relic features, and obtain the comprehensive degree of restoration of the cultural relic. S402. A fitting curve analysis was performed on the overall restoration degree at each time point during the restoration process to obtain the image restoration influence coefficient of the cultural relic restoration. The calculation process is as follows: ； In the above formula, X is the image restoration impact coefficient. This is to address the difference in color characteristics between the restored artifact and the original artifact. For the set standard feature difference, To address the differences in surface features between the restored artifact and the original artifact, The preset surface feature difference is defined by W1 and W2, which are weighting coefficients. I = {1, 2, 3, ..., N}, where N is the total number of feature points selected on the artifact for evaluation. F Y To determine the characteristics of the restored artifact, F S These are characteristics of primitive artifacts.

6. The image data verification system based on artificial intelligence according to claim 1, characterized in that, The specific process of constructing an image evolution recognition model is as follows: S500: Obtain the image feature dataset and the cultural relic image evolution feature coefficients, and combine them with the cultural relic image restoration influence coefficients to obtain the original image and the restored image of the cultural relic, which serve as the sample input data for the image evolution recognition model. The cultural relic image evolution feature coefficients and the cultural relic image restoration influence coefficients serve as the physical constraints of the image evolution recognition model. S501. Construct a neural network architecture based on physical constraints, and construct the loss function of the initial network model. Embed the equation residual gradient into the loss function to constrain the output of the image evolution recognition model. S502. Divide the image feature dataset into a training set and a test set in an 8:2 ratio. Adjust the number of hidden layer nodes according to the sample size of the training set, modify the last fully connected layer, and output the dynamic tracking results corresponding to the output dimension.

7. The image data verification system based on artificial intelligence according to claim 1, characterized in that, It also includes a feature analysis unit, which is used to identify the material characteristics of cultural relics and to classify and define the materials of cultural relics, including the following: S600. Based on the material characteristics of cultural relics, extract the image features of cultural relics, and assign initial weight values ​​to the cultural relic features of each material based on evolution analysis and the degree of influence of the restoration process. S601. Establish a weight adjustment feedback mechanism and conduct regular evaluation and adjustment of weight parameters in conjunction with historical restoration records of cultural relics.

8. The image data verification system based on artificial intelligence according to claim 1, characterized in that, Verify the consistency of the evolution time logic of real-time imagery and generate an overall consistency verification report. The specific process is as follows: S700: Obtain the real-time cultural relic image evolution tracking results, extract dynamic restoration features and compare them with the restoration features in the dynamic feature database to obtain the conformity of the current dynamic restoration features, and perform comparative verification based on the preset restoration conformity. S701. Obtain the deviation value between the real-time image features and the repair features in the dynamic feature database, and compare it with the preset threshold. If the deviation value is greater than the preset threshold, the repair process is determined to be abnormal, so as to generate a comparative analysis result. S702. Combining the time constraints of the cultural relic's edge, conduct a time matching degree analysis on the restoration features, verify the consistency of the evolution time logic of the real-time image, and generate an overall consistency verification report.

Images