Method and system for identifying electronic data of image file

Abstract

The invention discloses an image file electronic data identification method and system, and relates to the technical field of identification. Comprising the following steps: preprocessing input image file electronic data, including image denoising and gray scale equalization operation; preliminarily positioning a suspected stained area by using an edge detection algorithm; classifying the features of the suspected stained region; further segmenting and labeling the suspected stained regions through a region growing algorithm, extracting a fine region boundary, and generating structured labeling data for each stained region; and according to the structured annotation data and a classification result, grading each stained area, and providing a corresponding early warning strategy in combination with a grading result. According to the invention, the intelligence and practicability of image archive management are improved, the service life of archives is prolonged through scientific grading and early warning strategies, and the risk of information loss caused by contamination is reduced.

Description

Technical Field

[0001] The present invention relates to the field of recognition technology, in particular to a recognition method and system for electronic data of image archives. Background Art

[0002] With the rapid development of information and digital technology, the electronicization of image archives has become an important trend in the field of archive management. In modern image archive management, the digital storage and processing of image data have been widely used in many fields such as cultural protection, legal documents and historical records. However, during the long-term storage and use of image archives, they are often affected by external factors such as light, humidity, stains or wear, which inevitably lead to quality degradation. These problems include image blur, creases, oil stains and other forms of contamination. These contaminations not only affect the readability and authenticity of the archives, but may also lead to the loss of key information. Therefore, how to accurately and efficiently identify the contaminated areas of image archives, and classify, analyze and grade them has become an important direction for current research and development of related technologies.

[0003] At present, there are many technical solutions for the identification and processing of image archive contamination. For example, the automated recognition method based on image processing completes the identification of contaminated areas through edge detection, region segmentation or feature analysis. However, these methods still face many technical bottlenecks. On the one hand, the preprocessing algorithm of the existing technology has poor adaptability to the complex noise and non-uniform grayscale distribution of the image, which easily leads to inaccurate edge detection and regional positioning results. On the other hand, in the classification of contaminated areas, the analysis method of a single feature (such as color or texture) is difficult to cope with the diverse types of contamination and is prone to classification errors. In addition, the existing technology still has a lot of room for improvement in segmentation and annotation accuracy, especially the boundary extraction of small contaminated areas is relatively rough, which affects the reliability of subsequent classification and early warning. These shortcomings have limited the practicality and promotion and application of image archive contamination recognition to a certain extent. Summary of the invention

[0004] In view of the problems existing in the above-mentioned background technology, a method and system for identifying electronic data of image archives of the present invention are proposed.

[0005] Therefore, the problem to be solved by the present invention is how to effectively solve the problems of inaccurate positioning of damaged areas and low classification accuracy in the prior art by combining multi-dimensional feature fusion and classification algorithms, and improve the intelligence level of digital protection and restoration of image archives.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a method for identifying electronic data of image archives, which comprises: preprocessing the input electronic data of the image archives, including image denoising and grayscale equalization operations, and using the Canny edge detection algorithm to perform edge detection on the denoised and grayscale equalized image, extracting edge features in the image, and preliminarily locating the damaged area in the image through the edge features; based on the suspected damaged area, multi-dimensional feature fusion is performed using color, texture and morphological features, and the K-nearest neighbor algorithm K-NN is used to calculate the similarity between the points to be classified and the points in the training set using the Euclidean distance metric, k nearest neighbors are selected and voting decisions are made, and the features of the suspected damaged area are classified; the suspected damaged area is further segmented and labeled by a region growing algorithm, fine regional boundaries are extracted, and structured annotation data is generated for each damaged area; each damaged area is classified according to the structured annotation data and the classification results, and a corresponding early warning strategy is proposed in combination with the classification results.

[0007] As a preferred solution of the recognition method based on image archive electronic data of the present invention, the preliminary positioning of the suspected contaminated area includes the following steps: Perform gradient calculation and calculate the horizontal direction separately and vertical direction The gradient of , where x and y are the pixel coordinates of the image; And according to the horizontal direction and vertical direction The gradient of calculates the gradient magnitude and direction; Perform non-maximum suppression on the gradient amplitude, compare each pixel along the gradient direction, retain the local maximum value, and set other values ​​to zero; Use the double threshold method to set two thresholds: high threshold and low threshold , according to the double threshold, the gradient amplitude The gradient amplitude is binarized and the suspected contaminated area is preliminarily located.

[0008] As a preferred solution of the recognition method based on image archive electronic data of the present invention, wherein: the horizontal direction and vertical direction The gradient of is calculated as follows: ; The calculation process of the gradient amplitude and direction is as follows: ; ; in, is the gradient amplitude, For direction.

[0009] As a preferred solution of the recognition method based on image archive electronic data of the present invention, wherein: the gradient amplitude is set according to the double threshold value The binarization process of the gradient amplitude includes the following steps: if , then mark the pixel as a strong edge; if , then mark the pixel as non-edge; if , then determine whether it is an edge based on connectivity.

[0010] As a preferred solution of the recognition method based on image archive electronic data of the present invention, the further segmentation and marking of the suspected contaminated area by the region growing algorithm includes the following steps: selecting a number of seed points in the suspected contaminated area, setting a seed point set , where each is the initial point to be grown in the contaminated area; the region growing algorithm starts from the selected seed point and gradually checks the pixels adjacent to the current area, setting the similarity threshold of region growth as , if the gray value of the adjacent pixels The following conditions must be met: ; in, Pixel Gray value of Then the pixel The current growing region is added, and the neighboring pixels of the region continue to be checked. The region growing process is repeated until no more pixels can be added. After the region grows, the boundary of the contaminated region is extracted using gradient-based edge detection, and the boundary information is more accurate. Based on the growing region and the boundary information, structured annotation data is generated for each contaminated region.

[0011] As a preferred solution of the recognition method based on electronic data of image archives described in the present invention, the classification of each contaminated area includes: calculating the area ratio, grayscale deviation and morphological complexity in each contaminated area to quantify the degree of contamination; the morphological complexity is the shape complexity calculated by the perimeter and area of ​​the contaminated area. , the specific calculation formula is as follows: ; in, is the perimeter of the defaced area, is the area of ​​the defiled area; According to the area ratio, grayscale deviation and morphological complexity, the pollution degree classification rules are established.

[0012] In a second aspect, the present invention provides an identification system based on electronic data of image archives, which includes: a data preprocessing module, which is used to preprocess the input electronic data of the image archives, including image denoising and grayscale equalization operations; an edge positioning module, which is used to preliminarily locate the suspected contaminated area using an edge detection algorithm; a feature classification module, which is used to classify the features of the suspected contaminated area; a fine labeling module, which is used to further segment and label the suspected contaminated area through a region growing algorithm, extract fine region boundaries, and generate structured labeling data for each contaminated area; a contamination classification and warning module, which is used to classify each contaminated area according to the structured labeling data and classification results, and propose a corresponding warning strategy based on the classification results.

[0013] The beneficial effects of the present invention are: On the basis of preliminary positioning, the present invention combines color, texture and morphological features, and classifies suspected contaminated areas through multi-dimensional feature fusion, overcoming the problem of traditional classification methods relying too much on a single feature, achieving accurate classification of multiple types of contamination, reducing misclassification, and improving the reliability of labeled data; and refining the segmentation boundaries of contaminated areas, generating structured data, overcoming the problem of rough extraction of contamination boundaries in the prior art, and achieving refined processing of contaminated areas and standardized output of data. The present invention improves the intelligence and practicality of image archive management, extends the service life of archives through scientific classification and early warning strategies, and reduces the risk of information loss caused by contamination. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 The figure is a flow chart of the recognition method based on electronic data of image archives.

[0016] Figure 2 This is a structural diagram of the recognition system based on image archive electronic data. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0018] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0020] Example 1: Reference Figure 1~Figure 2 , which is the first embodiment of the present invention, provides a method for identifying electronic data of image archives, such as Figure 1 As shown, including: S1: Preprocess the input image file electronic data, including image denoising and grayscale equalization operations, and use edge detection algorithm to preliminarily locate the suspected contaminated area.

[0021] S1.1: Image denoising and grayscale equalization.

[0022] First, the input image file is denoised to reduce the interference of noise on subsequent processing, and then grayscale equalization is performed to optimize the brightness and contrast of the image to make the image features more prominent.

[0023] S1.2: Edge detection and preliminary region positioning.

[0024] The Canny edge detection algorithm is used to perform edge detection on the denoised and equalized image to extract the edge features in the image. Through the edge features, the damaged areas in the image, such as creases, oil stains or other worn areas, are preliminarily located.

[0025] Specifically, the Canny edge detection steps are as follows: First, perform Gaussian filtering on the input image to remove noise and obtain the filtered image. , x and y are the pixel coordinates of the image; The gradient of the filtered image is calculated. The commonly used gradient operator is the Sobel operator, which calculates the horizontal direction and vertical direction The gradient is: ; Furthermore, the gradient amplitude and direction It can be obtained by the following formula: ; ; In order to accurately locate the edge, it is necessary to suppress the non-maximum value of the gradient amplitude, compare each pixel along the gradient direction, retain the local maximum value, and set other values ​​to zero; Use the double threshold method to set two thresholds: high threshold and low threshold , according to the double threshold, the gradient amplitude And binarize the gradient magnitude: if , then mark the pixel as a strong edge; if , then mark the pixel as non-edge; if , then determine whether it is an edge based on connectivity.

[0026] Finally, based on the edge information obtained in the above steps, the suspected contaminated area is preliminarily located as a reference for subsequent area annotation.

[0027] S2: Classify the features of suspected damaged areas, such as creases, oil stains, wear, etc.

[0028] Based on the preliminary positioning, multi-dimensional feature fusion is performed using color, texture and morphological features. The damaged areas are initially classified based on the support vector machine (SVM) or K-nearest neighbor algorithm (K-NN), and the damaged areas are divided into types such as creases, oil stains, and wear, and the corresponding regional data is generated.

[0029] First, the color histogram of the region is calculated to describe the color distribution in the image.

[0030] Set the color components in each region to ,, calculate the color histogram H(C) as one of the features: ; in, is the color value of each pixel in the area, is the Dirac function, is the number of regions.

[0031] The local binary pattern (LBP) is used to extract the texture features of the region. For each pixel in the image region , the LBP value is obtained by comparing the neighborhood pixel value with the central pixel value: ; in, is the gray value of the center pixel, is the gray value of the adjacent pixel, S is the sign function, and P is the number of neighborhood pixels.

[0032] The morphological features of the contaminated area are extracted through morphological operations (such as erosion, dilation, opening and closing operations). These features are used to describe the shape, size and connectivity of the contaminated area's boundaries.

[0033] The above color, texture and morphological features are combined to construct a comprehensive feature vector , used for classification. Set the feature vector of each region to ,in, are the components of different features, and n is the number of feature vectors.

[0034] The K-nearest neighbor algorithm K-NN is used to calculate the similarity between the points to be classified and the training set points using the Euclidean distance metric, and the k nearest neighbors are selected and voted for decision.

[0035] According to the above processing results, the data of the contaminated area is automatically generated, including the category, bounding box and morphological characteristics of each contaminated area.

[0036] It should be noted that in the preliminary positioning of step S1, the boundary of the suspected contaminated area may be rough, but the texture, color and other features in the area can be extracted and used for classification. This feature extraction has low requirements on the absolute accuracy of the area boundary. If classification is performed after the completion of step S3, the computational complexity of the system will be increased and the classification results will be delayed, especially in scenarios that require real-time or efficient processing.

[0037] S3: The suspected contaminated areas are further segmented and labeled through the region growing algorithm, fine region boundaries are extracted, and structured annotation data is generated for each contaminated area.

[0038] Specifically, first, from the preliminary annotation results of the contaminated area obtained by edge detection and classification in step S1, select several seed points in the area. These seed points should be selected at the most obvious edge part, or at points with large grayscale or color changes in the area as starting points. Set the seed point set , where each is the initial point to be grown in the contaminated area.

[0039] Furthermore, the region growing algorithm starts from the selected seed point and gradually checks the pixels adjacent to the current region, and decides whether to add the pixel to the region based on the similarity between the pixel value (such as color, grayscale, texture features, etc.) and the pixels in the current region. The similarity threshold of region growing is set to , if the gray value of the adjacent pixels The following conditions must be met: ; in, Pixel The gray value of .

[0040] When the conditions are met, the pixel Join the current growing region Join the current growing region and continue to check the neighboring pixels of the region. The region growing process is repeated until there are no more pixels to join.

[0041] After region growing, gradient-based edge detection (such as the Sobel operator) is used to extract the boundary of the contaminated area to ensure more accurate boundary information. To avoid rough boundaries, boundary smoothing is performed, such as applying morphological dilation-erosion operations to obtain a smoothed boundary image.

[0042] Finally, structured annotation data is generated for each contaminated area based on the growth area and boundary information, including the following key features: area location coordinates, area area, area perimeter, and grayscale distribution (such as mean, standard deviation, etc.). The annotation information is output in a standard format.

[0043] S4: Based on the structured annotation data and classification results, each contaminated area is graded, and corresponding early warning strategies are proposed based on the classification results.

[0044] First, the following features are calculated from each soiled region to quantify the degree of soiling: Area ratio: the ratio of the defaced area to the total image area; Grayscale deviation: the grayscale mean difference between the contaminated area and its surrounding uncontaminated area; Morphological complexity: The shape complexity is calculated by the perimeter and area of ​​the stained area , the specific calculation formula is as follows: ; in, is the perimeter of the contaminated area. The closer it is to 1, the more regular the shape. The area of ​​the defiled area.

[0045] Based on the labeled data and classification results (such as the type of contamination) of the contaminated area, each area is comprehensively graded. The specific method is: Combine the contamination type with the classification indicators (such as area, grayscale deviation, morphological complexity, etc.) to establish classification rules for different types. For example, the crease area can focus on the morphological complexity, while the oil stain area focuses more on the area ratio.

[0046] According to the above indicators, establish the pollution degree classification rules, such as: Mild Defacement (Grade 1): Area ratio ; Grayscale deviation ; Morphological complexity .

[0047] Moderate stain (grade 2): Area ratio ; Grayscale deviation ; Morphological complexity .

[0048] Severe soiling (level 3): Area ratio ; Grayscale deviation ; The contamination is complex in shape or the area boundaries are irregular (high morphological complexity ).

[0049] According to the classification results, the early warning mechanism of the corresponding level is triggered: Mild warning: scheduled maintenance is recommended.

[0050] Moderate warning: It is recommended to speed up repairs or digital backup to avoid further deterioration.

[0051] Severe warning: Take protective measures immediately and give priority to repairs.

[0052] Finally, the classification and warning results will be output in a standardized electronic report format.

[0053] Furthermore, this embodiment also provides a recognition system based on electronic data of image archives, including: The data preprocessing module is used to preprocess the input electronic data of the image archive, including image denoising and grayscale equalization operations; An edge positioning module is used to preliminarily locate suspected contaminated areas using edge detection algorithms; A feature classification module, used to classify the features of the suspected contaminated area; A fine annotation module is used to further segment and annotate suspected contaminated areas through a region growing algorithm, extract fine region boundaries, and generate structured annotation data for each contaminated area; The pollution classification and early warning module is used to classify each pollution area according to the structured annotation data and classification results, and propose corresponding early warning strategies based on the classification results.

[0054] This embodiment also provides a computer device, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the recognition method based on image archive electronic data proposed in the above embodiment.

[0055] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0056] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for identifying electronic data based on image archives as proposed in the above embodiment is implemented.

[0057] In summary, based on the preliminary positioning, the present invention combines color, texture and morphological features, and classifies suspected contaminated areas through multi-dimensional feature fusion, overcoming the problem of traditional classification methods relying too much on a single feature, achieving accurate classification of multiple types of contamination, reducing misclassification, and improving the reliability of labeled data; and refining the segmentation boundaries of contaminated areas to generate structured data, overcoming the problem of rough extraction of contamination boundaries in the prior art, and achieving refined processing of contaminated areas and standardized output of data. The present invention improves the intelligence and practicality of image archive management, extends the service life of archives through scientific classification and early warning strategies, and reduces the risk of information loss due to contamination.

[0058] Example 2: Referring to Table 1 and Table 2, this is the second example of the present invention. This example provides a method for identifying electronic data in image archives. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation experiments.

[0059] The present invention selected 20 different types of image archives as test objects, covering various types such as historical documents, manuscript archives, electronic scans, etc. These archive data contain lightly, moderately and severely contaminated areas respectively.

[0060] The specific implementation steps of the test are as follows: Test environment preparation: A high-performance computer (processor: Intel Core i9-12900K, memory: 64 GB) was used to build the operating environment, and the image processing library OpenCV and the machine learning framework Scikit-Learn were installed. The resolution of the image files is 300 DPI, and the format is TIFF.

[0061] First, the input image data is preprocessed. The Gaussian filter algorithm is used to denoise the image to reduce the interference of noise on subsequent edge extraction; then the histogram equalization algorithm is used to optimize the image grayscale, improve the image contrast, and make the features of the stained area more prominent.

[0062] Next, the Canny edge detection algorithm is used to extract the edges of the processed image. Double thresholds are set to obtain the key edge features in the image, and the suspected contaminated areas are preliminarily located based on these features.

[0063] Based on the preliminary positioning, the color, texture and morphological features of the suspected contaminated area were extracted. The color histogram was divided into 10 color intervals, the texture features were extracted using the LBP algorithm (radius of 2, number of neighborhood pixels of 8), and the morphological features included the area and boundary shape complexity. Combined with the K-NN classification algorithm, the number of neighboring points was set, and the contaminated area was classified into three types: creases, oil stains and wear.

[0064] The region growing algorithm is used to select seed points from the suspected contaminated area (based on the principle of maximum grayscale difference), and the region is segmented based on the grayscale value similarity condition (for example, setting the threshold to 10). The region boundary is further extracted by the Sobel operator, and the boundary is smoothed by combining the expansion and erosion operations to generate structured annotation data including region location, area, grayscale distribution and other information.

[0065] Each contaminated area is graded based on quantitative indicators such as area ratio, grayscale deviation, and morphological complexity. Regular maintenance is recommended for lightly contaminated areas, repair suggestions are proposed for moderately contaminated areas, and emergency warning strategies for priority protection are proposed for heavily contaminated areas. Some experimental data are as follows: Table 1 Image processing data records Data file name Noise intensity after denoising (dB) Grayscale mean Grayscale standard deviation Number of detected edges Initial number of contaminated areas Canny detection time (ms) Archive_01 30.2 128 12.5 450 3 45 Archive_02 28.9 130 14.1 520 4 47 Archive_03 31.5 125 13.2 410 2 42 Archive_04 29.8 127 11.8 490 3 44 Archive_05 32.1 129 12.0 470 3 46 Table 2 Classification and grading results of contaminated areas Data file name Type of Defacement Area share (%) Grayscale deviation Morphological complexity Defect level Grading time (ms) Archive_01 Crease 5.2 8.5 1.12 Level 1 25 Archive_02 Oil pollution 12.4 25.6 1.45 Level 2 32 Archive_03 Wear 3.8 5.2 1.08 Level 1 22 Archive_04 Crease 15.3 50.1 1.75 Level 3 38 Archive_05 Oil pollution 7.6 12.4 1.25 Level 2 30 From the data in Table 1, it can be seen that the denoising and grayscale equalization steps in the present invention significantly reduce the noise intensity (an average reduction of 15 dB), while optimizing the grayscale mean and standard deviation of the image, enhancing the image contrast, and further improving the edge detection effect. For example, after the Archive_01 image is processed, 450 edge feature points are detected, while only about 300 can be detected in the unprocessed similar image. The increase in the number of edge extraction proves the effectiveness of the method of the present invention.

[0066] As can be seen from Table 2, the classification results of the contaminated areas classified by multi-dimensional feature fusion and K-NN algorithm have significant accuracy. For example, the comprehensive analysis of area proportion, grayscale deviation and morphological complexity realizes the reasonable division of light (level 1), moderate (level 2) and severe (level 3) contaminated areas. Taking Archive_04 as an example, its area proportion is as high as 15.3%, the grayscale deviation reaches 50.1, and it is accompanied by a high morphological complexity (1.75). It is classified as level 3 severe contamination, triggering an emergency warning strategy. Compared with existing technologies, this classification strategy can more accurately reflect the degree of contamination, thereby providing more precise protection measures.

[0067] In summary, the present invention achieves high-precision identification and intelligent management of damaged areas in image archives, has significant innovation and practicality compared with the existing technology, and provides a set of efficient solutions for the protection and management of image archives.

[0068] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for identifying electronic data of image archives, characterized in that: include: Preprocess the input electronic data of image archives, including image denoising and grayscale equalization operations, and use the Canny edge detection algorithm to perform edge detection on the denoised and grayscale equalized images to extract edge features in the image. Through the edge features, preliminarily locate the damaged area in the image; Based on the suspected contaminated area, multi-dimensional feature fusion is performed using color, texture and morphological features. The K-nearest neighbor algorithm K-NN is used to calculate the similarity between the points to be classified and the training set points using the Euclidean distance metric. The k nearest neighbors are selected and voted to classify the features of the suspected contaminated area. The suspected contaminated area is further segmented and labeled through the region growing algorithm, fine region boundaries are extracted, and structured annotation data is generated for each contaminated area; According to the structured annotation data and the classification results, each contaminated area is graded, and a corresponding early warning strategy is proposed based on the classification results.

2. The method for identifying electronic data based on image archives as claimed in claim 1, characterized in that: Initial location of suspected contamination areas includes the following steps: According to the filtered image Perform gradient calculation and calculate the horizontal direction separately and vertical direction The gradient of , where x and y are the pixel coordinates of the image; And according to the horizontal direction and vertical direction The gradient of calculates the gradient magnitude and direction; Perform non-maximum suppression on the gradient amplitude, compare each pixel along the gradient direction, retain the local maximum value, and set other values ​​to zero; Use the double threshold method to set two thresholds: high threshold and low threshold , according to the double threshold, the gradient amplitude The gradient amplitude is binarized and the suspected contaminated area is preliminarily located.

3. The method for identifying electronic data based on image archives as claimed in claim 2, characterized in that: The horizontal direction and vertical direction The gradient of is calculated as follows: ； The calculation process of the gradient amplitude and direction is as follows: ； ； in, is the gradient amplitude, For direction.

4. The method for identifying electronic data based on image archives as claimed in claim 3, characterized in that: The gradient amplitude is set according to the double threshold value The binarization process of the gradient amplitude includes the following steps: if , then mark the pixel as a strong edge; if , then mark the pixel as non-edge; if , then determine whether it is an edge based on connectivity.

5. The method for identifying electronic data based on image archives as claimed in claim 4, characterized in that: The further segmentation and labeling of the suspected contaminated area by the region growing algorithm comprises the following steps: Select several seed points in the suspected contaminated area and set the seed point set , where each is the initial point to be grown in the contaminated area; The region growing algorithm starts from the selected seed point and gradually checks the pixels adjacent to the current region, setting the similarity threshold of region growing as , if the gray value of the adjacent pixels The following conditions must be met: ； in, Pixel Gray value of Then the pixel Add the current growing region and continue to check the neighboring pixels of the region. The region growing process is repeated until there are no more pixels to add. After region growing, gradient-based edge detection is used to extract the boundary of the contaminated area, and the boundary information is more accurate; According to the growth area and the boundary information, structured annotation data is generated for each contaminated area.

6. The method for identifying electronic data based on image archives as claimed in claim 5, characterized in that: The classification of each defaced area includes: In each contaminated area, the area ratio, grayscale deviation and morphological complexity were calculated to quantify the degree of contamination; The morphological complexity is calculated by the perimeter and area of ​​the contaminated area. , the specific calculation formula is as follows: ； in, is the perimeter of the defaced area, is the area of ​​the defiled area; According to the area ratio, grayscale deviation and morphological complexity, the pollution degree classification rules are established.

7. A recognition system based on electronic data of image archives, based on the recognition method based on electronic data of image archives according to any one of claims 1 to 6, characterized in that: Also includes: The data preprocessing module is used to preprocess the input electronic data of the image archive, including image denoising and grayscale equalization operations; An edge positioning module is used to preliminarily locate suspected contaminated areas using edge detection algorithms; A feature classification module, used to classify the features of the suspected contaminated area; A fine annotation module is used to further segment and annotate suspected contaminated areas through a region growing algorithm, extract fine region boundaries, and generate structured annotation data for each contaminated area; The pollution classification and early warning module is used to classify each pollution area according to the structured annotation data and classification results, and propose corresponding early warning strategies based on the classification results.

Images