Target graph shielding processing method and system based on variable-focus wide-view-angle camera

Abstract

The invention relates to the technical field of image shielding processing, in particular to a target graph shielding processing method and system based on variable-focus wide-view-angle cameras, and the method comprises the steps: deploying a plurality of variable-focus cameras based on seat distribution, constructing a monitoring network, and collecting an initial image; identifying desktop features according to the images, establishing a unified coordinate system and mapping each desktop; when shooting is needed, querying candidate cameras according to the coordinate system, and determining a main camera according to an imaging area and a distance score; and performing quality evaluation on the main shot image, outputting the main shot image if the main shot image reaches the standard, otherwise, triggering a dynamic re-shooting mechanism. According to the invention, through cooperation of multiple cameras and intelligent visual angle selection, the problem that a single camera is easy to block is effectively solved, and automatic and high-quality image acquisition and processing are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image occlusion processing, and in particular to a target image occlusion processing method and system based on a variable-focus wide-view camera. BACKGROUND

[0002] Image acquisition and processing technology plays an increasingly important role in modern education scenarios, especially in applications such as online exams, homework submission, and classroom behavior analysis, which require clear and accurate image capture of specific student desks (such as test papers and workbooks). Traditional image acquisition methods usually rely on fixed monitoring cameras or manual handheld devices deployed in the classroom for shooting.

[0003] Currently, common classroom monitoring systems mostly use fixed-point deployment of wide-angle cameras or pan-tilt cameras. Wide-angle cameras have a large single-point field of view, which can cover a larger area, but have serious edge distortion and insufficient resolution of distant objects, and cannot flexibly adjust the viewing angle to avoid occlusion. Pan-tilt cameras can be rotated, but their physical movement is slow and cannot cover multiple areas simultaneously, making it difficult to meet the demand for rapid and concurrent image acquisition of multiple specific targets. When a specific student's desk needs to be shot, the above solutions are prone to cause the target to be partially or completely occluded due to the activities of other students, desk and chair occlusion, poor shooting angle, etc., and the obtained images cannot be used.

[0004] To solve the occlusion problem, existing technologies have adopted multiple camera solutions. However, these solutions simply increase the number of cameras to improve coverage, lacking coordinated management and intelligent scheduling of multi-camera resources. Human intervention is usually required, such as manually selecting an unoccluded camera view by an operator or manually controlling the pan-tilt to avoid obstacles. This approach is inefficient, slow to respond, and dependent on the operator's experience, making it difficult to achieve automated and real-time accurate image acquisition.

[0005] In addition, existing systems often lack the ability to automatically evaluate image quality. Whether the captured image is clear, complete, and usable usually requires manual post-checking, and cannot be judged and trigger a retake mechanism in real time during the acquisition phase. This may result in missing the best retake opportunity (such as during an exam) when image quality problems (such as blur, occlusion, distortion) are discovered, or requiring additional manpower and time for repeated acquisition.

[0006] Therefore, the existing technology has the problems of occlusion caused by fixed viewing angle and poor flexibility, low efficiency caused by lack of coordination of multi-camera systems, and inability to guarantee image usability caused by lack of automated quality evaluation and retake mechanism. SUMMARY

[0007] In order to overcome the problems of limited view angle, easy to be blocked, unstable image quality and low efficiency of manual adjustment when collecting images of specific desktop targets in complex environments such as classrooms in the prior art, the application provides a target graphic occlusion processing method and system based on a variable focal wide-angle camera.

[0008] The above application aims to achieve the following technical solutions: A target graphic occlusion processing method based on a variable focal wide-angle camera, the target graphic occlusion processing method based on a variable focal wide-angle camera comprises the following steps: Based on the student seat distribution in the classroom, a plurality of variable focal wide-angle camera devices are deployed to construct a monitoring field of view network of the monitoring area, and initial position images of each camera device are collected; According to the initial position image, the feature information of all student desktops is identified and extracted, the relative spatial position relationship between each camera device is calculated based on the feature information, and a unified coordinate system is established to map each desktop into the coordinate system; When a specific target desktop needs to be photographed, one or more candidate camera devices that can photograph the desktop are queried according to the coordinate system, and the optimal main camera device is determined among the candidate camera devices based on a preset picture scoring rule; An original image photographed by the main camera device is obtained, content analysis and image quality evaluation are performed on the original image, and an image quality score is obtained, when the image quality score meets the standard, a processed image is output, otherwise, a dynamic retake mechanism is triggered.

[0009] By using the above technical solutions, a plurality of variable focal wide-angle camera devices are adaptively deployed according to the student seat distribution in the classroom, thereby constructing a monitoring field of view network of the classroom without blind area, laying a hardware foundation for subsequent multi-view collaborative shooting, through collecting initial images and establishing a unified coordinate system, the desktops in the physical space are digitally mapped with the camera field of view, realizing global management of the shooting resources, when shooting is needed, the system can intelligently select the best view angle from multiple candidate cameras, effectively avoiding the occlusion, ensuring the success rate of the first shooting, the introduced image quality evaluation and dynamic retake mechanism constitute a closed-loop feedback system, ensuring that the final output image meets the predetermined standard in terms of content clarity, integrity and usability, greatly reducing the repeated manual operation caused by image quality problems, and improving the automation degree and reliability of the entire image collection process.

[0010] In a preferred example, the application can be further configured as follows: the plurality of variable focal wide-angle camera devices are deployed based on the student seat distribution in the classroom, a monitoring field of view network of the monitoring area is constructed, and initial position images of each camera device are collected, specifically comprising: According to the actual layout of the classroom and the row distribution of the student desktops, the installation position and number of the camera devices are determined to ensure that at least one desktop is covered by the field of view of one camera device, forming a monitoring area; An initial shooting angle and focal length are preset for each camera device, defined as an initial position, and an initial position image of all camera devices is acquired. Based on the acquired initial position image, desktop target detection is performed to identify all desktop areas in the initial position image, and the detected desktop areas are subjected to mask processing to obtain a processed initial position image.

[0011] By using the above technical solutions, the classroom layout is deployed in a targeted manner to ensure full coverage of the monitoring network from the source and eliminate blind spots. The initial position is preset and the image is acquired to provide a stable and consistent data source for subsequent feature extraction and coordinate system calibration. The image is subjected to desktop target detection and mask processing to effectively filter background interference, highlight the key target (desktop), significantly improve the accuracy and efficiency of subsequent feature matching, and provide pure data input for accurately establishing a spatial mapping relationship.

[0012] In a preferred example, the application can be further configured to: identify and extract feature information of all student desktops based on the initial position image, calculate the relative spatial position relationship between the camera devices based on the feature information, and establish a unified coordinate system to map each desktop into the coordinate system, specifically including: Feature point extraction is performed on the processed initial position image of each camera device using a scale-invariant feature transform algorithm to obtain multiple sets of feature point description data. The multiple sets of feature point description data are matched two by two to find corresponding feature points belonging to the same desktop between different images. According to the successfully matched feature point pairs, a homography matrix between adjacent camera device images is calculated to derive the relative pose and distance between the camera devices. Taking a fixed corner point of the classroom as the origin, a unified coordinate system is constructed in combination with the relative pose and distance between the camera devices, and the field of view range of each desktop and each camera device is mapped into the coordinate system to form a desktop-camera device mapping relationship database.

[0013] By adopting the technical scheme, the feature extraction algorithm with scale invariance such as SIFT is adopted, so that stable and reliable feature points can be extracted under different visual angles and different spatial levels, and the spatial relative relationship between the cameras can be solved with high precision through feature matching and homography matrix calculation, which is the core of realizing multi-view cooperation and intelligent angle selection, a unified coordinate system is constructed, and a mapping relationship database is formed, which is equivalent to creating a "digital twin" virtual monitoring space, so that the system can globally and quantitatively perceive and schedule the cameras and desktop targets in the physical space, and provide important spatial information support for subsequent intelligent decision-making.

[0014] In a preferred example, the application can be further configured to: the one or more candidate cameras that can shoot the desktop are queried according to the coordinate system, and the optimal main camera is determined among the candidate cameras based on a preset picture scoring rule, specifically including: All candidate cameras that can shoot the target desktop are determined according to the coordinate system, the imaging area proportion of the target desktop in the picture shot by each candidate camera is calculated, and an imaging area score is formed; The Euclidean distance between the center point of the target desktop and the optical center of each candidate camera in the coordinate system is calculated, and a distance score is formed; According to the imaging area score and the distance score, all candidate cameras are weighted and sorted, and the camera with the highest sorting is selected as the main camera.

[0015] By adopting the technical scheme, the mapping database established in advance is queried, all possible cameras covering the target can be quickly and accurately locked, the decision-making efficiency is improved, the imaging area (directly affecting the size and clarity of the target in the picture) and the shooting distance (affecting the perspective distortion and image details) two key factors are comprehensively considered for weighted scoring, so that the selection decision of the main camera is more scientific and comprehensive, and the theoretically best imaging effect can be actively selected from multiple feasible schemes, so that high-quality original images are preferentially obtained, the image quality is improved from the source, and the dependence on subsequent image enhancement or re-shooting is reduced.

[0016] In a preferred example, the application can be further configured to: acquire the original image shot by the main camera, perform content analysis and image quality evaluation on the original image, and obtain an image quality score, specifically including: Paper target detection is performed on the original image, if no paper is detected or the area proportion of the paper region is too low, it is directly determined that the shooting fails; When the paper is detected in the original image, instance segmentation is performed on the detected paper region, the edge contour thereof is extracted, and the concave features of the contour shape are obtained; performing human body part detection on the original image to determine whether the paper edge is adhered to a human body part, and generating an adhesion feature; calculating the sharpness of a text edge and the number of detected characters in the paper region, and generating a paper text feature based on the sharpness of the text edge and the number of characters; generating an image quality score based on the concave feature, the adhesion feature, and the paper text feature.

[0017] By adopting the technical solutions described above, a multi-dimensional and multi-level image quality evaluation system is constructed. First, it is determined whether the target exists (paper), and then it is finely evaluated from three dimensions of physical form (concave representation of folding or bending), shielding condition (adhesion representing human shielding), and content quality (text sharpness and number representing clarity and information integrity). This evaluation method closely matches the core requirements of image quality in actual application scenarios such as examination and homework collection, and can comprehensively and accurately quantify the usability of the image, providing an objective and reliable decision basis for triggering the retake mechanism, and avoiding subjective misjudgment.

[0018] In a preferred example, the application can be further configured to trigger a dynamic retake mechanism, specifically including: When the image quality score is lower than the qualified threshold, it is determined that the shooting effect is poor, the coordinate system is queried, and if the target desktop is only mapped to one candidate camera device, the main camera device is controlled to adjust the focal length and angle for retaking after a certain interval of time; If the target desktop is mapped to multiple candidate camera devices in the coordinate system, the weighted and sorted candidate camera device group is called, and the suboptimal camera device is selected for retaking; The newly taken image is obtained, and the newly taken image and the original image are compared in terms of quality score, and the image with the highest quality score after comparison is selected as the output result.

[0019] By adopting the technical solutions described above, a differentiated retake plan is developed according to the available camera resources: parameter adjustment is attempted when there is a single camera, and the alternative view is switched decisively when there are multiple cameras, thereby improving the success rate and efficiency of retaking. Finally, the optimal image is automatically selected for output through quality comparison, ensuring the reliability of the final result. The entire process does not require human intervention, forming a complete "shooting-evaluation-decision-retake-output" automated closed loop, which significantly improves the robustness and user experience of the system.

[0020] In a preferred example, the application can be further configured to, before the output processed image, the target image occlusion processing method based on the variable focal width angle camera further includes: The high-quality image is obtained according to the quality image score, the paper profile is extracted based on the high-quality image, the perspective transformation matrix of the paper profile is calculated, the paper area in the selected high-quality image is corrected by perspective, and a distortion-free front view is obtained; The front view is subjected to text line direction detection, and is subjected to rotation processing so that the text line is in a horizontal direction; The front view is subjected to contrast and brightness adjustment based on an adaptive histogram equalization algorithm, and the contrast of the paper background and the text area is enhanced. The processed front view is cropped into a standard document picture containing only a target paper area and is output and stored in parallel.

[0021] By adopting the technical solutions, a series of post-processing optimization is performed on the images that pass the evaluation, the standardization degree and the usability of the images are further improved, the perspective correction eliminates the geometric distortion caused by the shooting angle, the document is restored to a front view, the text line is horizontally rotated so that the document content conforms to the reading habit, the subsequent manual review or OCR recognition is facilitated, the adaptive histogram equalization optimizes the contrast of the image, and the legibility of the text is enhanced, especially under the condition of uneven illumination, the effect is remarkable, and finally the standardized document picture is cropped and output, so that the output result is uniform in specification, pure in background and prominent in content, and the requirements of subsequent archiving, recognition, auditing and other application links for the quality of the input image are greatly met.

[0022] In the second aspect, the above-mentioned application purpose is achieved by the following technical solutions: A target graphic occlusion processing system based on a variable-focus wide-view camera, comprising: A camera arrangement module, configured to arrange a plurality of variable-focus wide-view cameras based on the distribution of student seats in a classroom, construct a monitoring field of view network of the monitoring area, and collect initial position images of each camera; A coordinate system construction module, configured to identify and extract feature information of all student desks according to the initial position images, calculate the relative spatial position relationship between each camera based on the feature information, and establish a unified coordinate system to map each desk to the coordinate system; A main camera determination module, configured to identify and extract feature information of all student desks according to the initial position images, calculate the relative spatial position relationship between each camera based on the feature information, and establish a unified coordinate system to map each desk to the coordinate system; An image quality evaluation module, configured to obtain an original image captured by the main camera, perform content analysis and image quality evaluation on the original image, obtain an image quality score, output a processed image when the image quality score meets the standard, and otherwise trigger a dynamic retake mechanism.

[0023] By adopting the technical scheme, multiple variable focal wide-view cameras are adaptively arranged in the classroom according to the seat distribution of students, so that a monitoring field of view network without blind area in the classroom is constructed, which lays a hardware foundation for subsequent multi-view collaborative shooting, initial images are collected and a unified coordinate system is established, the desktop in the physical space and the camera field of view are digitally mapped, global management of shooting resources is realized, when shooting is needed, the system can intelligently select the best view angle from multiple candidate cameras, effectively avoiding the obstruction, ensuring the success rate of the first shooting, the introduced image quality evaluation and dynamic re-shooting mechanism constitute a closed-loop feedback system, ensuring that the final output image meets the predetermined standards in content clarity, integrity and usability, greatly reducing repeated manual operations caused by image quality problems, and improving the automation degree and reliability of the entire image acquisition process.

[0024] In a third aspect, the above object of the present application is achieved by the following technical scheme: An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above target graphic occlusion processing method based on the variable focal wide-view camera when executing the computer program.

[0025] In a fourth aspect, the above object of the present application is achieved by the following technical scheme: A computer readable storage medium stores a computer program, and the computer program implements the steps of the above target graphic occlusion processing method based on the variable focal wide-view camera when executed by a processor.

[0026] In summary, the present application includes at least one of the following beneficial technical effects: 1、In the classroom, multiple variable focal wide-view cameras are adaptively arranged according to the seat distribution of students, so that a monitoring field of view network without blind area in the classroom is constructed, which lays a hardware foundation for subsequent multi-view collaborative shooting, initial images are collected and a unified coordinate system is established, the desktop in the physical space and the camera field of view are digitally mapped, global management of shooting resources is realized, when shooting is needed, the system can intelligently select the best view angle from multiple candidate cameras, effectively avoiding the obstruction, ensuring the success rate of the first shooting, the introduced image quality evaluation and dynamic re-shooting mechanism constitute a closed-loop feedback system, ensuring that the final output image meets the predetermined standards in content clarity, integrity and usability, greatly reducing repeated manual operations caused by image quality problems, and improving the automation degree and reliability of the entire image acquisition process; 2. Based on the pre-established mapping database, the query can quickly and accurately lock all possible target coverage cameras, improve decision efficiency, comprehensively consider the imaging area (directly affecting the size and clarity of the target in the picture) and shooting distance (affecting perspective distortion and image details) two key factors for weighted scoring, making the selection decision of the main camera more scientific and comprehensive, and actively selecting the best imaging effect in theory from multiple feasible schemes, so as to preferentially obtain high-quality original images, improve image quality from the source, and reduce the dependence on subsequent image enhancement or re-shooting; 3. Using SIFT and other scale-invariant feature extraction algorithms ensures that stable and reliable feature points can be extracted under different angles, different spatial levels, etc. Through feature matching and homography matrix calculation, the spatial relative relationship between cameras can be accurately solved, which is the core of multi-angle cooperation and intelligent angle selection. A unified coordinate system is constructed and a mapping relationship database is formed, which is equivalent to creating a "digital twin" virtual monitoring space, so that the system can globally and quantitatively perceive and dispatch the cameras and desktop targets in the physical space, providing important spatial information support for subsequent intelligent decision-making; 4. Quickly detect whether the target (paper) exists, perform preliminary screening, and then perform fine evaluation from three dimensions of physical form (concave representation of folding or bending), shielding condition (adhesion representing human shielding), and content quality (text sharpness and quantity representing clarity and information integrity). This evaluation method closely matches the actual application scenarios such as exams and homework collection, and can comprehensively and accurately quantify the usability of images, providing objective and reliable decision-making basis for triggering the re-shooting mechanism, avoiding subjective misjudgment. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a flowchart of a target image shielding processing method based on a variable focal wide-angle camera in an embodiment of the present application; Figure 2 is an implementation flowchart of step S10 in the target image shielding processing method based on a variable focal wide-angle camera in an embodiment of the present application; Figure 3 is an implementation flowchart of step S20 in the target image shielding processing method based on a variable focal wide-angle camera in an embodiment of the present application; Figure 4 is an implementation flowchart of step S30 in the target image shielding processing method based on a variable focal wide-angle camera in an embodiment of the present application; Figure 5 is an implementation flowchart of step S40 in the target image shielding processing method based on a variable focal wide-angle camera in an embodiment of the present application; Figure 6 is another implementation flowchart of step S40 in the target graphic occlusion processing method based on a variable-focus wide-view camera in an embodiment of the present application; Figure 7 is another implementation flowchart of the target graphic occlusion processing method based on a variable-focus wide-view camera in an embodiment of the present application; Figure 8 is a principle block diagram of the target graphic occlusion processing system based on a variable-focus wide-view camera in an embodiment of the present application; Figure 9 is a schematic diagram of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0028] The present application will be further described in detail below with reference to the accompanying drawings.

[0029] In an embodiment, as shown in Figure 1 The present application discloses a target graphic occlusion processing method based on a variable-focus wide-view camera, which specifically comprises the following steps: S10: Based on the distribution of student seats in the classroom, a plurality of variable-focus wide-view cameras are deployed, a monitoring field of view network of the monitoring area is constructed, and an initial position image of each camera is collected.

[0030] Specifically, in a classroom environment, according to the actual arrangement pattern of student seats, such as the number of rows and columns, the spacing, etc., a plurality of cameras with zoom and wide-view functions are reasonably deployed on the top or side wall of the classroom, ensuring that each student's desktop is covered by the field of view of at least one camera, forming a monitoring area without blind area, and thus forming a complete monitoring field of view network.

[0031] Further, according to a preset initial shooting angle and focal length of each camera, i.e., the initial position, an image is collected as an initial position image, providing a data basis for subsequent coordinate system construction and desktop mapping.

[0032] S20: According to the initial position image, the feature information of all student desktops is recognized and extracted, the relative spatial position relationship between the cameras is calculated based on the feature information, and a unified coordinate system is established, mapping each desktop into the coordinate system.

[0033] Specifically, a target detection and feature extraction algorithm is used to process the initial position image, recognize all student desktop areas in the image, and extract the feature information of the desktops, such as the desktop corners, textures, and identifiers, etc. Based on the desktop feature information, the relative pose and distance relationship between the cameras are derived through feature point matching and geometric transformation.

[0034] Further, a unified spatial coordinate system is established with a fixed corner of the classroom, such as the upper left corner of the classroom, as the origin, and the spatial positions of each desktop and the field of view range of each camera are mapped into the coordinate system, forming a desktop-camera mapping relationship database.

[0035] S30: When a specific target desktop needs to be photographed, one or more candidate cameras that can photograph the desktop are queried according to the coordinate system, and the optimal main camera is determined among the candidate cameras based on a preset picture scoring rule.

[0036] Specifically, when the system receives a photographing request for a target desktop, all candidate cameras that can cover the desktop are queried according to the established coordinate system, the photographed desktop image is scored for each candidate camera, and the camera with the highest score is selected as the main camera according to the preset picture scoring rule, thereby ensuring the optimal photographing angle and the best image quality.

[0037] S40: Obtain the original image photographed by the main camera, perform content analysis and image quality evaluation on the original image, obtain an image quality score, output the processed image when the image quality score meets the standard, otherwise trigger the dynamic retake mechanism.

[0038] Specifically, the main camera is controlled to photograph the target desktop to obtain an original image, and in-depth content analysis is performed on the image, including paper target detection, paper region instance segmentation, human part detection, and text sharpness calculation, thereby evaluating the image quality, such as whether it is clear, whether there is occlusion, and the degree of distortion, and forming an image quality score.

[0039] Further, when the image quality score meets the qualified threshold, subsequent processing is performed on the image, including perspective correction, rotation, and enhancement, and the processed image is output, and if the image quality score does not meet the qualified threshold, the dynamic retake mechanism is triggered until a qualified image is obtained.

[0040] In this embodiment,

[0041] In an embodiment, as shown in FIG. 1, in step S10, a plurality of variable focal wide-angle cameras are deployed based on the seat distribution of students in the classroom, a monitoring field of view network of the monitoring area is constructed, and an initial position image of each camera is collected, specifically including: Figure 2 S11: According to the actual layout of the classroom and the row and column distribution of the student desktops, the installation position and number of the cameras are determined to ensure that any desktop is covered by the field of view of at least one camera, forming a monitoring area.

[0042] ​Specifically, by on-site measurement of the classroom layout or design paper analysis, the specific row and column distribution and spacing of the student desktops are determined, and according to the field of view coverage requirements such as avoiding obstruction and reducing distortion, the installation position, height, angle and number of the camera devices are determined by using field of view simulation or geometric calculation to ensure that there is no monitoring blind area and a comprehensive monitoring network is formed.

[0043] S12: An initial shooting angle and focal length are preset for each camera device, which is defined as an initial position, and an initial position image of all camera devices is obtained.

[0044] Specifically, a standard initial parameter is set for each camera device, such as adjusting the focal length to the widest angle and the angle to view the desktop from above, and an initial position image at this time is collected. The initial position image aims to provide consistent data sources for subsequent feature extraction and coordinate system construction.

[0045] S13: Based on the collected initial position image, desktop target detection is performed to identify all desktop areas in the initial position image, and the detected desktop areas are subjected to mask processing to obtain a processed initial position image.

[0046] Specifically, a deep learning target detection algorithm is used to identify the desktop areas in the initial position image, and the identified desktop areas are subjected to binary mask processing to highlight the desktop areas in the image and filter out background interference, facilitating subsequent feature extraction and matching.

[0047] In an embodiment, as shown in Figure 3 In step S20, the feature information of all student desktops is identified and extracted according to the initial position image, the relative spatial position relationship between the camera devices is calculated based on the feature information, and a unified coordinate system is established to map each desktop into the coordinate system, including: S21: Feature point extraction is performed on the processed initial position images of each camera device using a scale-invariant feature transform algorithm to obtain multiple sets of feature point description data.

[0048] Specifically, the SIFT algorithm is used to detect feature points and extract feature descriptors from the images of each camera device after mask processing. The SIFT algorithm has scale invariance and rotation invariance, can effectively extract stable feature points of the desktop, and generate feature description vectors to form multiple sets of feature point data such as table corners and specific markers on the desktop.

[0049] S22: The multiple sets of feature point description data are matched two by two to find corresponding feature points belonging to the same desktop in different images.

[0050] Specifically, a feature matching algorithm is used to match different image feature point description data pairwise. By calculating the similarity between each group of feature points, matching feature point pairs are found, and mismatched feature points are identified through cross-validation to ensure the accuracy of feature point matching.

[0051] S23: Based on the successfully matched feature point pairs, calculate the homography matrix between adjacent camera device images, and deduce the relative pose and distance between each camera device.

[0052] Specifically, using successfully matched feature point pairs, the RANSAC algorithm is used to calculate the homography matrix between adjacent camera device images. By decomposing the homography matrix and combining it with the built-in parameters of the camera devices, the relative pose and distance relationship between the camera devices are derived, providing geometric constraints for coordinate system construction.

[0053] S24: Using a fixed corner point in the classroom as the origin, and combining the relative poses and distances between each camera device, a unified coordinate system is constructed, and the field of view of each desktop and each camera device is mapped into the coordinate system to form a database of desktop-camera device mapping relationships.

[0054] Specifically, a fixed corner point in the classroom is selected as the origin of the coordinate system. Combining the relative pose and distance relationship between the camera devices obtained in step S23, the corner coordinates of each desktop and the optical center coordinates of each camera device are transformed to a unified coordinate system through coordinate transformation. The field of view of each camera device is calculated, such as the projection of the visual cone in the coordinate system. A mapping relationship table between the desktop and the camera device is established and stored in the database for subsequent queries.

[0055] In one embodiment, such as Figure 4 As shown, in step S30, one or more candidate camera devices that can capture images of the desktop are queried according to the coordinate system, and the optimal main camera device is determined from the candidate camera devices based on preset image scoring rules. This specifically includes: S31: Determine all candidate camera devices that can capture images of the target desktop based on the coordinate system, calculate the proportion of the target desktop's image area in the images captured by each candidate camera device, and form an image area score.

[0056] Specifically, using the established coordinate system, all camera devices covering the target desktop are quickly queried as candidates. For each candidate camera device, based on its internal parameters and pose, the proportion of the target desktop's projected area on the imaging plane of each candidate camera device to the total screen area is calculated as the imaging area score. The higher the proportion, the higher the score, indicating that the desktop is larger in the screen.

[0057] S32: Calculate the Euclidean distance between the target desktop center point and the optical center of each candidate camera in the coordinate system, and form a distance score.

[0058] Specifically, in the coordinate system, the Euclidean distance between the coordinates of the target desktop center point and the coordinates of the optical center of each candidate camera is calculated as a distance score. The closer the distance, the better the potential image quality, and the higher the score.

[0059] S33: According to the imaging area score and the distance score, all candidate cameras are weighted and sorted, and the camera with the highest score is selected as the main camera.

[0060] Specifically, the obtained imaging area score and distance score are normalized, and according to the set weight value, such as an imaging area weight of 0.6 and a distance weight of 0.4, the score of the image captured by each candidate camera is calculated, and the cameras are sorted from high to low according to the score. The camera with the highest score is selected as the main camera for shooting.

[0061] In an embodiment, as shown in Figure 5 In step S40, the original image captured by the main camera is obtained, and content analysis and image quality evaluation are performed on the original image to obtain an image quality score, which specifically includes: S41: Paper target detection is performed on the original image, and if no paper is detected or the area of the paper region is too small, the shooting is directly determined to fail.

[0062] Specifically, target detection is used to detect paper in the original image captured by the main camera. If no corresponding paper region is detected or the area of the paper region accounts for less than a threshold of the total image area, it is considered that the paper is missing or too small, and the current shooting is directly determined to fail, triggering a re-shooting mechanism.

[0063] S42: When paper is detected in the original image, instance segmentation is performed on the detected paper region, the edge contour is extracted, and the concave feature of the contour shape is obtained.

[0064] Specifically, when paper is detected, an instance segmentation model is further used to accurately segment the paper pixel region, the edge contour of the paper region is extracted, and the concave degree of the contour shape is analyzed, such as contour convex hull defect detection. If the concave is too large, it may indicate that the paper is folded or occluded, which affects the picture quality score as a concave feature.

[0065] S43: Human body part detection is performed on the original image to determine whether the paper edge is adhered to the human body part, and an adhesion feature is generated.

[0066] Specifically, a human key point detection model is used to detect human body parts in the image, such as arms and hands, and to determine whether the paper edge contour overlaps or adheres to the pixels of the human body parts. If adhesion exists, it may obscure the content of the paper and is recorded as a negative feature, thus obtaining the adhesion feature.

[0067] S44: Calculate the sharpness of the text edges within the paper area and the number of detected characters, and generate paper text features based on the text edge sharpness and the number of characters.

[0068] Specifically, an edge sharpness algorithm is applied to the detected paper area to evaluate the clarity of the text within the paper area. At the same time, OCR is used to recognize and count the number of characters to form paper text features. The sharper the text edges and the more characters recognized, the higher the text feature score, which represents better image quality.

[0069] S45: Generate an image quality score based on the indentation features, adhesion features, and paper text features.

[0070] Specifically, features such as the degree of indentation, adhesion, text sharpness, and number of characters are assigned weights, and a weighted composite score is calculated as the image quality score. A passing threshold is set; if the score is higher than the threshold, the image quality is considered acceptable; otherwise, it is considered unacceptable.

[0071] In one embodiment, such as Figure 6 As shown, in step S40, the dynamic replay mechanism is triggered, which specifically includes: S46: When the image quality score is lower than the qualified threshold, it is determined that the shooting effect is poor. The coordinate system is queried. If the target desktop is only mapped to one candidate camera device, the main camera device is controlled again to adjust the focal length and angle for reshooting after a certain interval.

[0072] Specifically, if the quality score is unsatisfactory and the mapping relationship shows that only one candidate camera device covers the target desktop, the system will wait for a short time (e.g., 2 seconds) and then control the same camera device to fine-tune the focus or shooting angle (e.g., slightly zoom in or adjust the tilt angle) to try to take a picture again to obtain a better image.

[0073] S47: If the target desktop is mapped to multiple candidate camera devices in the coordinate system, the weighted and sorted candidate camera device group is invoked, and the second-best camera device is selected for reshooting.

[0074] Specifically, if there are multiple candidate devices, the second-best device with the second-highest overall score after the main camera device is selected as the new shooting device according to the ranking result of step S30, and it is controlled to shoot to obtain images from different perspectives.

[0075] S48: Obtain a newly taken image, compare the newly taken image with the original image in terms of quality score, and select the image with the highest quality score after comparison as the output result.

[0076] Specifically, the same quality evaluation process is performed on the newly taken image to obtain a quality score. The score of the newly taken image is compared with the score of the original image, and the image with the highest score is selected as the final output result to ensure the optimal quality of the output image.

[0077] In an embodiment, as shown in FIG. 8, before outputting the processed image, the target graphic occlusion processing method based on the variable focal wide-angle camera further includes: Figure 7 S50: Obtain a high-quality image according to the quality image score, extract a paper outline based on the high-quality image, calculate a perspective transformation matrix of the paper outline, perform perspective correction on the paper region in the selected high-quality image, and obtain a distortion-free front view.

[0078] Specifically, the image with the highest quality score is selected, the perspective transformation matrix is calculated using the accurate outline points (such as four corner points) of the paper region, the matrix is applied to perform perspective correction on the paper region, the tilted and distorted view is converted into a standard front rectangular view, and the perspective distortion is eliminated.

[0079] S60: Perform text line direction detection on the front view and perform rotation processing to make the text line horizontal.

[0080] Specifically, the text line angle detection is performed on the corrected front view, the inclination angle of the text line is calculated, and the image is rotated to make the text line horizontal, which facilitates reading and subsequent processing.

[0081] S70: Perform contrast and brightness adjustment on the front view based on an adaptive histogram equalization algorithm to enhance the contrast between the paper background and the text region.

[0082] Specifically, the adaptive histogram equalization algorithm is applied to enhance the contrast of the image. This algorithm can effectively improve the local contrast, highlight the difference between the text region and the paper background, and improve the image clarity and readability.

[0083] S80: Crop the processed front view into a standard document picture containing only the target paper region and output and store in parallel.

[0084] Specifically, the enhanced image is cropped into a standard size picture containing only the target paper according to the accurate bounding box of the paper region. The final processed standard document picture is output to the display interface and archived, and the entire processing process is completed.

[0085] ​It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0086] In one embodiment, a target image occlusion processing system based on a variable focal length field-of-view camera is provided. This system corresponds one-to-one with the target image occlusion processing method based on a variable focal length field-of-view camera described in the above embodiments. For example... Figure 8 As shown, the target image occlusion processing system based on a variable focal width angle camera includes a camera device deployment module, a coordinate system construction module, a main camera device determination module, and an image quality evaluation module. Detailed descriptions of each functional module are as follows: The camera deployment module is used to deploy multiple variable focal width angle cameras based on the distribution of student seats in the classroom, construct a monitoring field of view network for the monitoring area, and acquire initial position images of each camera. The coordinate system construction module is used to identify and extract feature information of all student desktops based on the initial position image, calculate the relative spatial position relationship between each camera device based on the feature information, and establish a unified coordinate system to map each desktop into the coordinate system. The main camera device determination module is used to identify and extract feature information of all student desktops based on the initial position image, calculate the relative spatial position relationship between each camera device based on the feature information, and establish a unified coordinate system to map each desktop into the coordinate system. The image quality assessment module is used to acquire the original image captured by the main camera device, perform content analysis and image quality assessment on the original image to obtain an image quality score. When the image quality score meets the standard, the processed image is output; otherwise, a dynamic reshoot mechanism is triggered.

[0087] Specific limitations regarding the target image occlusion processing system based on a variable focal length wide-angle camera can be found in the limitations of the target image occlusion processing method based on a variable focal length wide-angle camera mentioned above, and will not be repeated here. Each module in the aforementioned target image occlusion processing system based on a variable focal length wide-angle camera can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in hardware or independently of the processor in an electronic device, or stored in software in the memory of the electronic device, so that the processor can call and execute the corresponding operations of each module.

[0088] In one embodiment, an electronic device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9The electronic device includes a processor, a memory, a network interface and a database connected through a system bus. The processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the electronic device is configured to store monitoring data of multiple-source patients and a diabetes analysis model. The network interface of the electronic device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement a target graphic occlusion processing method based on a variable focal wide-view camera.

[0089] In one embodiment, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the following steps when executing the computer program: A plurality of variable focal wide-view cameras are deployed based on the seat distribution of students in the classroom to construct a monitoring field of view network of the monitoring area, and an initial position image of each camera is collected; According to the initial position image, feature information of all student desks is identified and extracted, a relative spatial position relationship between the cameras is calculated based on the feature information, and a unified coordinate system is established to map each desk into the coordinate system; When a specific target desk needs to be photographed, one or more candidate cameras that can photograph the desk are queried according to the coordinate system, and the optimal main camera is determined among the candidate cameras based on a preset picture scoring rule; An original image photographed by the main camera is obtained, content analysis and image quality evaluation are performed on the original image to obtain an image quality score, and when the image quality score meets the standard, a processed image is output, otherwise, a dynamic retake mechanism is triggered.

[0090] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the following steps: A plurality of variable focal wide-view cameras are deployed based on the seat distribution of students in the classroom to construct a monitoring field of view network of the monitoring area, and an initial position image of each camera is collected; According to the initial position image, feature information of all student desks is identified and extracted, a relative spatial position relationship between the cameras is calculated based on the feature information, and a unified coordinate system is established to map each desk into the coordinate system; When a specific target desktop needs to be photographed, one or more candidate camera devices that can photograph the desktop are queried according to the coordinate system, and a main camera device is determined from the candidate camera devices based on a preset picture score rule. An original image photographed by the main camera device is acquired, content analysis and image quality evaluation are performed on the original image, an image quality score is obtained, when the image quality score meets a standard, a processed image is output, otherwise, a dynamic retake mechanism is triggered.

[0091] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0092] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the above-mentioned functions.

[0093] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for processing target image occlusion based on a variable focal width angle camera, characterized in that, The target image occlusion processing method based on a variable focal width angle camera includes the following steps: Based on the distribution of student seats in the classroom, multiple variable focal width angle cameras are deployed to construct a monitoring field of view network for the monitoring area, and the initial position images of each camera are collected. Based on the initial position image, feature information of all student desktops is identified and extracted. The relative spatial position relationship between each camera device is calculated based on the feature information, and a unified coordinate system is established to map each desktop into the coordinate system. When it is necessary to photograph a specific target desktop, one or more candidate camera devices that can photograph the desktop are queried according to the coordinate system, and the optimal main camera device is determined from the candidate camera devices based on the preset image scoring rules. The system acquires the original image captured by the main camera device, performs content analysis and image quality assessment on the original image to obtain an image quality score. If the image quality score meets the standard, the processed image is output; otherwise, a dynamic reshoot mechanism is triggered.

2. The target image occlusion processing method based on a variable focal width angle camera according to claim 1, characterized in that, The method involves deploying multiple variable focal length cameras based on the student seating arrangement in the classroom to construct a monitoring field of view network for the monitored area, and acquiring initial position images of each camera. Specifically, this includes: Based on the actual layout of the classroom and the row and column distribution of the students' desks, determine the installation location and number of camera devices to ensure that any desk is covered by the field of view of at least one camera device, forming a monitoring area; For each camera device, an initial shooting angle and focal length are preset and defined as the initial position. The initial position images of all camera devices are then acquired. Desktop target detection is performed based on the acquired initial position image. All desktop regions in the initial position image are identified, and the detected desktop regions are masked to obtain the processed initial position image.

3. The target image occlusion processing method based on a variable focal width angle camera according to claim 1, characterized in that, The process of identifying and extracting feature information from all student desktops based on the initial position image, calculating the relative spatial position relationships between each camera device based on the feature information, and establishing a unified coordinate system to map each desktop into the coordinate system specifically includes: The scale-invariant feature transform algorithm was used to extract feature points from the processed initial bit images of each camera device to obtain multiple sets of feature point description data. The multiple sets of feature point description data are matched pairwise to find the corresponding feature points belonging to the same desktop in different images; Based on the successfully matched feature point pairs, the homography matrix between adjacent camera device images is calculated, and the relative pose and distance between each camera device are derived. Using a fixed corner point in the classroom as the origin, and combining the relative poses and distances between each camera device, a unified coordinate system is constructed. The field of view of each desktop and each camera device is then mapped into the coordinate system, forming a database of desktop-camera device mapping relationships.

4. The target image occlusion processing method based on a variable focal width angle camera according to claim 1, characterized in that, The step of querying one or more candidate camera devices that can capture images of the desktop according to the coordinate system, and determining the optimal main camera device from the candidate camera devices based on preset image scoring rules, specifically includes: Based on the coordinate system, all candidate camera devices capable of capturing images of the target desktop are determined, and the proportion of the target desktop's image area in the images captured by each candidate camera device is calculated to form an image area score. Calculate the Euclidean distance between the center point of the target desktop and the optical center of each candidate camera device in the coordinate system to form a distance score; Based on the imaging area score and distance score, all candidate camera devices are weighted and ranked, and the camera device with the highest ranking is selected as the main camera device.

5. The target image occlusion processing method based on a variable focal width angle camera according to claim 1, characterized in that, The process of acquiring the original image captured by the main camera device, performing content analysis and image quality assessment on the original image to obtain an image quality score specifically includes: The original image is subjected to paper target detection. If no paper is detected or the paper area accounts for too low a percentage of the image, the shooting is directly determined to be a failure. When paper is detected in the original image, the detected paper area is segmented, its edge contour is extracted, and the concave features of the contour shape are obtained. Human body part detection is performed on the original image to determine whether the paper edge is stuck to the human body part, and an adhesion feature is generated. Calculate the sharpness of the text edges within the paper area and the number of detected characters, and generate paper text features based on the text edge sharpness and the number of characters; An image quality score is generated based on the indentation features, adhesion features, and paper text features.

6. The target image occlusion processing method based on a variable focal width angle camera according to claim 1, characterized in that, The dynamic replay triggering mechanism specifically includes: When the image quality score is lower than the qualified threshold, it is determined that the shooting effect is poor. The coordinate system is queried. If the target desktop is only mapped to one candidate camera device, the main camera device is controlled again to adjust the focus and angle for reshooting after a certain interval. If the target desktop is mapped to multiple candidate camera devices in the coordinate system, the weighted and sorted candidate camera device group is invoked, and the second-best camera device is selected for reshooting. Acquire newly captured images, compare the quality scores of the newly captured images with those of the original images, and select the image with the highest quality score as the output.

7. The target image occlusion processing method based on a variable focal width angle camera according to claim 1, characterized in that, Before outputting the processed image, the target image occlusion processing method based on a variable focal width angle camera further includes: High-quality images are obtained based on quality image scores. Paper outlines are extracted from the high-quality images. The perspective transformation matrix of the paper outlines is calculated. Perspective correction is performed on the paper areas in the selected high-quality images to obtain a distortion-free front view. The text line direction of the front view is detected and rotated to make the text line horizontal. The contrast and brightness of the front view are adjusted based on the adaptive histogram equalization algorithm to enhance the contrast between the paper background and the text area. The processed front view is cropped into a standard document image containing only the target paper area and then output and stored in parallel.

8. A target image occlusion processing system based on a variable focal width angle camera, characterized in that, The target image occlusion processing system based on a variable focal width angle camera includes: The camera deployment module is used to deploy multiple variable focal width angle cameras based on the distribution of student seats in the classroom, construct a monitoring field of view network for the monitoring area, and acquire initial position images of each camera. The coordinate system construction module is used to identify and extract feature information of all student desktops based on the initial position image, calculate the relative spatial position relationship between each camera device based on the feature information, and establish a unified coordinate system to map each desktop into the coordinate system. The main camera device determination module is used to identify and extract feature information of all student desktops based on the initial position image, calculate the relative spatial position relationship between each camera device based on the feature information, and establish a unified coordinate system to map each desktop into the coordinate system. The image quality assessment module is used to acquire the original image captured by the main camera device, perform content analysis and image quality assessment on the original image to obtain an image quality score. When the image quality score meets the standard, the processed image is output; otherwise, a dynamic reshoot mechanism is triggered.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the target image occlusion processing method based on a variable focal width angle camera as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the target image occlusion processing method based on a variable focal width angle camera as described in any one of claims 1 to 7.

Images