An artificial intelligence-based picture generation method and system

Abstract

The present application relates to the field of image processing, and more particularly to a picture generation method and system based on artificial intelligence, the method comprising: detecting a feature object distribution state corresponding to a target real scene reference image, and determining an image category of the target real scene reference image according to the feature object distribution state; respectively performing combination correlation detection on each edge distribution category image, and determining a preselected edge combination according to a combination correlation degree of a to-be-combined image and the target edge distribution category image; determining an effective edge combination according to a feature similarity corresponding to each preselected edge combination; determining an effective center image according to a feature difference degree of the center distribution category image; respectively transmitting the effective center image, the effective edge image and an invalid supplementary image to corresponding image classification storage ends, and allowing a user to select the effective center image and the effective edge image to perform image splicing generation; and the present application effectively divides the image to be spliced, thereby improving the image splicing generation efficiency.

Description

Technical Field

[0001] This invention relates to the field of image processing, and in particular to an image generation method and system based on artificial intelligence. Background Technology

[0002] Panoramic images are gaining increasing attention due to their immersive experience. In practical applications, user demands for images often exceed the field of view of ordinary cameras, making high-definition, wide-angle images a necessity. To reduce costs and obtain high-quality images, image stitching technology has emerged, and its applications are becoming increasingly widespread, including geological surveys, power line inspections, weather monitoring, space exploration, medical diagnosis, and virtual simulation, demonstrating broad development prospects and high research value. Panoramic image stitching specifically involves registering and transforming multiple small-sized images from the same scene to create a high-quality, larger panoramic image. However, in actual image stitching generation, limitations in user shooting capabilities often lead to overlap between the original images used for stitching or repeated shooting of the same scene. Therefore, how to effectively filter and select initial images for image stitching generation to improve efficiency is a problem urgently needing to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN106530214A discloses an image stitching system and an image stitching method. The method includes: acquiring multiple images corresponding to multiple scene points using multiple image acquisition devices, each image corresponding to a scene point including a reference image and an image to be stitched, wherein the reference image and the image to be stitched have overlapping areas; extracting multiple candidate feature point pairs between each reference image and the image to be stitched; removing redundant feature point pairs from the extracted candidate feature point pairs to obtain stitching feature point pairs; estimating the rotation matrix and offset matrix between the multiple image acquisition devices using the stitching feature point pairs; and stitching the images corresponding to the scene points according to the rotation matrix and offset matrix to obtain a stitched image. It is evident that the above technical solution has the following problems: it requires high-quality shooting; if there are many repeated shots or similar images to be stitched, it is impossible to effectively select the images to be stitched, resulting in poor image stitching generation efficiency. Summary of the Invention

[0004] To address this issue, the present invention provides an image generation method and system based on artificial intelligence, which overcomes the problem in the prior art that, when there are many images to be stitched together that are repeatedly taken or are similar, the inability to effectively select the images to be stitched together results in poor image stitching generation efficiency.

[0005] To achieve the above objectives, the present invention provides an image generation method based on artificial intelligence, comprising:

[0006] Detect the distribution state of features corresponding to the target real-scene reference image, and determine the image category of the target real-scene reference image based on the distribution state of features;

[0007] For each edge distribution category image, a combination association detection is performed, and the pre-selected edge combination is determined based on the combination association degree between the image to be combined and the target edge distribution category image.

[0008] Valid edge combinations are determined based on the feature similarity corresponding to each pre-selected edge combination;

[0009] Valid center images are determined based on the feature differences of the center distribution category images;

[0010] The effective center image, effective edge image, and invalid supplementary image are transmitted to the corresponding image classification storage terminal respectively. The user selects the effective center image and effective edge image to generate the image by stitching.

[0011] Furthermore, for each target real-scene reference image, the distribution state of the features corresponding to the target real-scene reference image is detected to determine the image category of the target real-scene reference image, including:

[0012] The central distribution category image is a target real-world reference image whose corresponding feature distribution state is in the first preset feature distribution state.

[0013] The edge distribution category image is a target real-world reference image whose corresponding feature distribution state is in the second preset feature distribution state.

[0014] Furthermore, the distribution state of the features includes:

[0015] The first preset feature distribution state is that the feature distribution density corresponding to the target real scene reference image is less than the preset feature distribution density.

[0016] The second preset feature distribution state is that the feature distribution density corresponding to the target real scene reference image is greater than or equal to the preset feature distribution density.

[0017] If there are no features in the target real-world reference image, it is recorded as an invalid supplementary image.

[0018] Furthermore, when performing combination association detection on images of each edge distribution category, the image to be combined with the highest combination association degree with the target edge distribution category image is detected. If the combination association degree between the image to be combined and the target edge distribution category image is greater than the preset combination association degree, the two edge distribution category images are determined to be successfully combined and recorded as a pre-selected edge combination.

[0019] In this context, the edge distribution category image to be combined and detected is denoted as the target edge distribution category image, and other edge distribution category images are denoted as the images to be combined.

[0020] Furthermore, the feature similarity of each pre-selected edge combination is detected through iteration, and the effective edge combination is determined based on the feature similarity.

[0021] If the feature similarity is within the first preset feature similarity range, then the two pre-selected edge combinations are determined to be valid edge combinations;

[0022] If the feature similarity is within the second preset feature similarity range, then the effective edge combination is determined based on the feature loss.

[0023] Furthermore, the feature loss degree corresponding to the two pre-selected edge combinations is detected. The feature loss degree of a single pre-selected edge combination is determined by detecting the minimum area of ​​the combined edge features in the edge distribution category image of the pre-selected edge combination, and the minimum area of ​​the combined edge features is recorded as the feature loss degree corresponding to the pre-selected edge combination.

[0024] The pre-selected edge combination with the smallest corresponding feature loss among the two pre-selected edge combinations is denoted as the effective edge combination.

[0025] Furthermore, the feature difference of the detection center distribution category images is traversed, and the effective center images are determined based on the feature difference.

[0026] If the feature difference of the centrally distributed category image is within the first preset feature difference range, then the centrally distributed category image is determined to be an invalid supplementary image.

[0027] If the feature difference of the center distribution category image is within the second preset feature difference range, then the center distribution category image is determined to be a valid center image.

[0028] Furthermore, the effective center image, effective edge image, and invalid supplementary image are transmitted to the corresponding image classification storage terminals respectively.

[0029] This invention also provides an artificial intelligence-based image generation system, comprising:

[0030] The image segmentation unit is used to determine the image category of the target real-world reference image based on the distribution state of the features corresponding to the target real-world reference image at the user input terminal.

[0031] The first processing unit, which is connected to the image segmentation unit, is used to perform combination association detection on images of each edge distribution category, determine pre-selected edge combinations based on the combination association degree between the image to be combined and the target edge distribution category image, and determine effective edge combinations based on the feature similarity corresponding to each pre-selected edge combination.

[0032] The second processing unit, which is connected to the image segmentation unit, is used to determine the effective center image based on the feature difference degree of the center distribution category image.

[0033] The data transmission unit is connected to the image segmentation unit, the first processing unit, and the second processing unit, respectively, and is used to transmit the target real-scene reference image from the user terminal to the image segmentation unit and to transmit the effective center image, effective edge image, and invalid supplementary image to the corresponding image classification storage terminal.

[0034] Compared with the prior art, the beneficial effects of the present invention are that, in the technical solution of the present invention, the image category of the target real-scene reference image is determined according to the feature distribution state corresponding to each target real-scene reference image, so that the initial image is effectively divided. Moreover, the division process is based on the feature distribution state, which reflects the distribution state of features within the target real-scene reference image, thereby making the image category division of the target real-scene reference image more consistent with the actual working scenario, and thus improving the speed of the subsequent judgment process.

[0035] Furthermore, in this invention, combined association detection is performed on images of each edge distribution category and pre-selected edge combinations are obtained accordingly. The combination association degree reflects whether the features of the image to be combined and the target edge distribution category image after stitching match the actual scene. Subsequently, the feature similarity corresponding to each pre-selected edge combination is detected through traversal, and the effective edge combination is determined based on the feature similarity. This avoids the situation where there are many repeated shots or similar images to be stitched, which leads to poor efficiency in subsequent image stitching generation.

[0036] Furthermore, in this invention, if the feature similarity is within the second preset feature similarity range, then the effective edge combination is determined based on the feature loss degree. The feature loss degree reflects the degree of influence of the stitching seam on the feature graphic after stitching, thereby making the selected effective edge combination more in line with the image stitching requirements, thus improving the image stitching efficiency of this invention, while reducing the requirements for the user's shooting ability.

[0037] Furthermore, the effective center image, effective edge image, and invalid supplementary image are transmitted to the corresponding image classification storage terminal, enabling users to effectively select the images to be generated by image stitching, reducing the requirements on the user's shooting ability, and improving the image stitching speed of the present invention. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of an image generation method based on artificial intelligence according to an embodiment of the present invention;

[0039] Figure 2This is a schematic diagram illustrating the detection of feature distribution status corresponding to a target real-scene reference image to determine the image category of the target real-scene reference image according to an embodiment of the present invention;

[0040] Figure 3 This is a schematic diagram illustrating how an embodiment of the present invention determines effective edge combinations based on feature similarity;

[0041] Figure 4 This is a schematic diagram of an artificial intelligence-based image generation system according to an embodiment of the present invention. Detailed Implementation

[0042] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0043] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0044] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0045] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0046] Please see Figure 1 As shown, it is a schematic diagram of an image generation method based on artificial intelligence according to an embodiment of the present invention. The present invention provides an image generation method based on artificial intelligence, including:

[0047] Detect the distribution state of features corresponding to the target real-scene reference image, and determine the image category of the target real-scene reference image based on the distribution state of features;

[0048] For each edge distribution category image, a combination association detection is performed, and the pre-selected edge combination is determined based on the combination association degree between the image to be combined and the target edge distribution category image.

[0049] Valid edge combinations are determined based on the feature similarity corresponding to each pre-selected edge combination;

[0050] Valid center images are determined based on the feature differences of the center distribution category images;

[0051] The effective center image, effective edge image, and invalid supplementary image are transmitted to the corresponding image classification storage terminal respectively. The user selects the effective center image and effective edge image to generate the image by stitching.

[0052] Please see Figure 2 As shown, this is a schematic diagram illustrating how the distribution state of features corresponding to a target real-scene reference image is detected to determine the image category of the target real-scene reference image according to an embodiment of the present invention. For each target real-scene reference image, detecting the distribution state of features corresponding to the target real-scene reference image to determine the image category includes:

[0053] The central distribution category image is a target real-world reference image whose corresponding feature distribution state is in the first preset feature distribution state.

[0054] The edge distribution category image is a target real-world reference image whose corresponding feature distribution state is in the second preset feature distribution state.

[0055] Please continue reading. Figure 1 As shown, the distribution state of the features includes:

[0056] The first preset feature distribution state is that the feature distribution density corresponding to the target real scene reference image is less than the preset feature distribution density.

[0057] The second preset feature distribution state is that the feature distribution density corresponding to the target real scene reference image is greater than or equal to the preset feature distribution density.

[0058] If there are no features in the target real-world reference image, it is recorded as an invalid supplementary image.

[0059] The method for confirming the feature distribution density is as follows: taking the center point of the target real-scene reference image as the center, a circle including all features in the image is constructed, and the area of ​​the circle is recorded as the feature distribution density. The preset value of the feature distribution density can be set by the user in the actual working scenario. That is, the preset feature distribution density reflects whether the feature is at the edge of the image. Those skilled in the art will understand that image stitching will cause quality problems in the image effect of the feature at the stitching point. Therefore, a preset feature distribution density value is provided, which is 50% of the area of ​​the target real-scene reference image.

[0060] The methods for identifying features include:

[0061] Image preprocessing involves performing preprocessing on the image, including noise reduction, contrast enhancement, and size normalization.

[0062] Feature extraction involves using deep learning networks to extract key features from an image, including edges, corners, and textures.

[0063] Feature matching involves matching extracted features with features of known objects, or using a classifier to classify features, thereby determining the features in the target real-world reference image.

[0064] This is content that is easily understood by those skilled in the art, and will not be elaborated upon here.

[0065] Specifically, when performing combination association detection on images of each edge distribution category, the image to be combined with the highest combination association degree with the target edge distribution category image is detected. If the combination association degree between the image to be combined and the target edge distribution category image is greater than the preset combination association degree, the two edge distribution category images are determined to be successfully combined and recorded as a pre-selected edge combination.

[0066] In this context, the edge distribution category image to be combined and detected is denoted as the target edge distribution category image, and other edge distribution category images are denoted as the images to be combined.

[0067] Specifically, the method for confirming the combination correlation is as follows: The image to be combined is stitched together with any two sides of the target edge distribution category image, and the proportion of similar feature points at the stitching point is detected. This proportion is denoted as the combination correlation. Similar feature points are pixels with the same color at the same height as the feature points at the stitching point of the image to be combined and the target edge distribution category image. The proportion of similar feature points = similar feature points / (number of pixels of the feature points at the stitching point of the image to be combined + number of pixels of the feature points at the stitching point of the target edge distribution category image). The corresponding height is the height relative to the bottom edge of the image with the stitching seam as the height direction. A preset value for the combination correlation is provided, which can be set by the user according to the actual working scenario. The greater the user's requirement for feature stitching accuracy, the larger the preset value for the combination correlation. A preset value for the combination correlation is provided: Preset combination correlation = (number of pixels of the feature points at the stitching point of the image to be combined + number of pixels of the feature points at the stitching point of the target edge distribution category image) / 2.

[0068] Please see Figure 3 As shown, this is a schematic diagram of determining effective edge combinations based on feature similarity in an embodiment of the present invention. It iterates through and detects the feature similarity corresponding to each pre-selected edge combination, and determines the effective edge combination based on feature similarity.

[0069] If the feature similarity is within the first preset feature similarity range, then the two pre-selected edge combinations are determined to be valid edge combinations;

[0070] If the feature similarity is within the second preset feature similarity range, then the effective edge combination is determined based on the feature loss.

[0071] Specifically, for a single pre-selected edge combination, the feature similarity is determined by detecting the similarity between the stitched images of the two images in the pre-selected edge combination and the images after other pre-selected edge combinations. The maximum image similarity corresponding to the pre-selected edge combination is recorded as the feature similarity corresponding to the pre-selected edge combination. The image similarity is obtained using the SSIM algorithm, with a numerical range of [-1, 1,]. The SSIM algorithm is a metric for measuring the similarity between two images, measuring image similarity from three aspects: brightness, contrast, and structure. The steps of the SSIM algorithm include:

[0072] Calculate μ, σ², and σXY. μ is the average of all pixel values ​​in the two images, representing the image brightness. σ² is the average of the squares of the differences between the pixel values ​​of the two images and their mean, representing the image contrast. σXY is the covariance, which is the average of the products of the differences between corresponding pixel values ​​in the two images.

[0073] The SSIM index is calculated by combining similarity measures of brightness, contrast, and structure to form the SSIM index function. The three corresponding functions μ, σ2, and σXY are weighted and combined to obtain an SSIM value between -1 and 1. This is a concept that is easily understood by those skilled in the art and will not be elaborated here.

[0074] The values ​​within the first preset feature similarity range are all less than 0, and the values ​​within the second preset feature similarity range are all greater than or equal to 0.

[0075] Please continue reading. Figure 1 As shown, the feature loss degree corresponding to two pre-selected edge combinations is detected. The feature loss degree of a single pre-selected edge combination is determined by detecting the minimum area of ​​the combined edge features in the edge distribution category image of the pre-selected edge combination, and the minimum area of ​​the combined edge features is recorded as the feature loss degree corresponding to the pre-selected edge combination.

[0076] The pre-selected edge combination with the smallest corresponding feature loss among the two pre-selected edge combinations is denoted as the effective edge combination.

[0077] Specifically, the feature difference of the detection center distribution category images is traversed, and the effective center images are determined based on the feature difference.

[0078] If the feature difference of the centrally distributed category image is within the first preset feature difference range, then the centrally distributed category image is determined to be an invalid supplementary image.

[0079] If the feature difference of the center distribution category image is within the second preset feature difference range, then the center distribution category image is determined to be a valid center image.

[0080] The method for confirming the feature difference degree of a single center distribution category image is as follows: detect the feature difference characterization value of the center distribution category image with other center distribution category images, and record the corresponding minimum feature difference characterization value as the feature difference degree of the center distribution category image. Feature difference characterization value = |(M1-M2)| / M1+|(S1-S2)| / S1, where M1 is the number of features in the center distribution category image, M2 is the number of features in the center distribution category image compared with the center distribution category image, S1 is the area of ​​the largest feature in the center distribution category image, and S2 is the area of ​​the largest feature in the center distribution category image compared with the center distribution category image. The preset feature difference degree range can be set by the user according to the actual working scenario. The greater the user's requirement for image stitching accuracy, the smaller the maximum value in the first preset feature difference degree range. A preset feature difference degree range is provided, where the values ​​in the first preset feature difference degree range are all less than 20%, and the values ​​in the second preset feature difference degree range are all greater than or equal to 20%.

[0081] Specifically, the effective center image, effective edge image, and invalid supplementary image are transmitted to the corresponding image classification storage terminals respectively.

[0082] The image classification storage terminal includes a first classification storage terminal, a second classification storage terminal, and a third classification storage terminal. The first classification storage terminal stores valid center images, the second classification storage terminal stores valid edge images, and the third classification storage terminal stores invalid supplementary images. During actual image stitching, users can select images from the first, second, and third classification storage terminals according to the actual working scenario, thereby improving the image stitching generation rate and reducing the impact of poor image stitching effect caused by repeated shooting or a large number of similar images to be stitched.

[0083] Please see Figure 4 As shown, this is a schematic diagram of an artificial intelligence-based image generation system according to an embodiment of the present invention. The present invention also provides an artificial intelligence-based image generation system, comprising:

[0084] The image segmentation unit is used to determine the image category of the target real-world reference image based on the distribution state of the features corresponding to the target real-world reference image at the user input terminal.

[0085] The first processing unit, which is connected to the image segmentation unit, is used to perform combination association detection on images of each edge distribution category, determine pre-selected edge combinations based on the combination association degree between the image to be combined and the target edge distribution category image, and determine effective edge combinations based on the feature similarity corresponding to each pre-selected edge combination.

[0086] The second processing unit, which is connected to the image segmentation unit, is used to determine the effective center image based on the feature difference degree of the center distribution category image.

[0087] The data transmission unit is connected to the image segmentation unit, the first processing unit, and the second processing unit, respectively, and is used to transmit the target real-scene reference image from the user terminal to the image segmentation unit and to transmit the effective center image, effective edge image, and invalid supplementary image to the corresponding image classification storage terminal.

[0088] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0089] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based picture generation method, characterized by, include: The distribution state of features corresponding to the target real-scene reference image is detected, and the image category of the target real-scene reference image is determined based on the distribution state of features, including edge distribution category images and center distribution category images; For each edge distribution category image, a combination association detection is performed, and the pre-selected edge combination is determined based on the combination association degree between the image to be combined and the target edge distribution category image. The effective edge combination is determined based on the feature similarity corresponding to each pre-selected edge combination. Specifically, the feature similarity corresponding to each pre-selected edge combination is detected by traversing the process, and the effective edge combination is determined based on the feature similarity. If the feature similarity is within the first preset feature similarity range, then the two pre-selected edge combinations are determined to be valid edge combinations; If the feature similarity is within the second preset feature similarity range, then the effective edge combination is determined based on the feature loss. Valid center images are determined based on the feature differences of the center distribution category images; The effective center image, effective edge image, and invalid supplementary image are respectively transmitted to the corresponding image classification storage terminal, and the user selects the effective center image and effective edge image for image stitching. For a single pre-selected edge combination, the feature similarity is determined by detecting the similarity between the images after the two images in the pre-selected edge combination are stitched together and the images after other pre-selected edge combinations are stitched together. The maximum image similarity corresponding to the pre-selected edge combination is recorded as the feature similarity corresponding to the pre-selected edge combination.

2. The image generation method based on artificial intelligence according to claim 1, characterized in that, For each target real-scene reference image, the distribution state of the corresponding features in the target real-scene reference image is detected to determine the image category of the target real-scene reference image, including: The central distribution category image is a target real-world reference image whose corresponding feature distribution state is in the first preset feature distribution state. The edge distribution category image is a target real-world reference image whose corresponding feature distribution state is in the second preset feature distribution state. 3.The AI-based picture generating method of claim 2, wherein, The distribution state of the feature includes: The first preset feature distribution state is that the feature distribution density corresponding to the target real scene reference image is less than the preset feature distribution density. The second preset feature distribution state is that the feature distribution density corresponding to the target real scene reference image is greater than or equal to the preset feature distribution density. If there are no features in the target real-world reference image, it is recorded as an invalid supplementary image. 4.The AI-based picture generating method of claim 2, wherein, When performing combination association detection on images of each edge distribution category, the image to be combined with the highest combination association degree with the target edge distribution category image is detected. If the combination association degree between the image to be combined and the target edge distribution category image is greater than the preset combination association degree, the two edge distribution category images are determined to be successfully combined and recorded as a pre-selected edge combination. In this context, the edge distribution category image to be combined and detected is denoted as the target edge distribution category image, and other edge distribution category images are denoted as the images to be combined. 5.The AI-based picture generating method of claim 4, wherein, The feature loss degree corresponding to two pre-selected edge combinations is detected. The feature loss degree of a single pre-selected edge combination is determined by detecting the minimum area of ​​the combined edge features in the edge distribution category image of the pre-selected edge combination, and the minimum area of ​​the combined edge features is recorded as the feature loss degree corresponding to the pre-selected edge combination. 6.The AI-based picture generation method of claim 5, wherein, The pre-selected edge combination with the smallest corresponding feature loss among the two pre-selected edge combinations is denoted as the effective edge combination.

7. The artificial intelligence-based picture generation method of claim 6, wherein, Traverse the feature difference scores of the detection center distribution category images and determine the effective center images based on the feature difference scores; If the feature difference of the centrally distributed category image is within the first preset feature difference range, then the centrally distributed category image is determined to be an invalid supplementary image. If the feature difference of the center distribution category image is within the second preset feature difference range, then the center distribution category image is determined to be a valid center image. 8.The AI-based picture generating method of claim 7, wherein, The valid center image, valid edge image, and invalid supplementary image are transmitted to the corresponding image classification storage terminals respectively.

9. An artificial intelligence-based picture generation system applying the method according to any one of claims 1 to 8, characterized in that, include: The image segmentation unit is used to determine the image category of the target real-world reference image based on the distribution state of the features corresponding to the target real-world reference image at the user input terminal. The first processing unit, which is connected to the image segmentation unit, is used to perform combination association detection on images of each edge distribution category, determine pre-selected edge combinations based on the combination association degree between the image to be combined and the target edge distribution category image, and determine effective edge combinations based on the feature similarity corresponding to each pre-selected edge combination. The second processing unit, which is connected to the image segmentation unit, is used to determine the effective center image based on the feature difference degree of the center distribution category image. The data transmission unit is connected to the image segmentation unit, the first processing unit, and the second processing unit, respectively, and is used to transmit the target real-scene reference image from the user terminal to the image segmentation unit and to transmit the effective center image, effective edge image, and invalid supplementary image to the corresponding image classification storage terminal.

Images