An intelligent panoramic splicing video control system

Abstract

The present invention discloses an intelligent panoramic stitching video control system, which relates to the field of video surveillance, including: a labeling module, which is used to receive surveillance videos and perform frame labeling on dynamic targets in the surveillance videos; a tracing module, which is used to select the framed and labeled dynamic targets in the surveillance videos and perform tracing on the dynamic targets; an enhancement module, which is used to frame select areas in the surveillance videos as local enhancement targets and perform enhancement processing on the local enhancement targets; the present invention identifies and labels dynamic targets based on a background difference method, and can also divide the video screen to clarify the regional source device, while intercepting the dynamic target area image and storing it according to the similarity; after the dynamic target is selected, it can be compared with the stored image, sorted by the source timestamp and determined in combination with the region to achieve tracing; key areas such as access control can be framed and enhanced, and after segmentation, they can be displayed on a split screen on a mobile device, and at the same time have a risk target prediction function.

Description

Technical Field

[0001] The present invention relates to the technical field of video surveillance, and in particular to an intelligent panoramic splicing video control system. Background Art

[0002] Panoramic video surveillance stitches video footage from multiple perspectives into a single panoramic image, achieving comprehensive surveillance. Advanced algorithms process image edges and blend them, eliminating stitching gaps and color variations. This system is widely used in large settings such as plazas and train stations, providing a comprehensive, continuous field of view and improving security efficiency and management.

[0003] The invention patent application with application number 202310490718.2 discloses an intelligent mobile panoramic video image control system, comprising a guide base having an annular guide groove formed thereon, in which a slider is slidably connected. A monitoring compartment is fixedly provided at the lower end of the slider, wherein a monitoring module is provided in the monitoring compartment, a fill light module is provided on the monitoring compartment for cooperating with the monitoring module, a light-transmitting glass for cooperating with the monitoring module is rotatably provided on the lower side of the monitoring compartment, a wireless cleaning module is provided on the monitoring compartment for cooperating with the light-transmitting glass, and a movement control mechanism is provided on the guide base for cooperating with the monitoring compartment. The application aims to solve the problem that "currently, the position of the surveillance cameras used in nursing home bedrooms is generally fixed, and thus can only perform single-point monitoring, so that only scene information of a single viewpoint in a certain observation direction can be obtained, the field of view is narrow, and only a limited scene can be monitored. Moreover, during long-term use, the surveillance cameras are easily adhered to by dust, thereby affecting the monitoring effect of the surveillance cameras. Moreover, since the surveillance cameras are generally installed at high places, it is inconvenient for staff to clean the dust on the surveillance camera lenses."

[0004] However, for panoramic monitoring of public places, due to the large volume of dynamic targets in panoramic monitoring videos, although panoramic monitoring can achieve global monitoring of the monitoring area, the amount of information in panoramic monitoring videos is too large to cooperate with monitoring managers to implement effective monitoring management.

[0005] To this end, we propose an intelligent panoramic stitching video control system. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the prior art, the present invention provides an intelligent panoramic splicing video control system, which can effectively solve the problems of the prior art.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0008] The present invention discloses an intelligent panoramic splicing video control system, comprising:

[0009] The labeling module is used to receive surveillance videos and frame the dynamic targets in the surveillance videos; the tracing module is used to select the dynamic targets framed and marked in the surveillance videos and trace the dynamic targets; the enhancement module is used to frame the selected areas in the surveillance videos as local enhanced targets and perform enhancement processing on the local enhanced targets; the ROI display module is used to receive the framed areas that have been enhanced in the enhancement module and display the framed areas independently; the judgment module is used to judge whether there is a framed and marked dynamic target selected by the tracing module in the historical video of the framed area independently displayed in the ROI display module. When the judgment result is yes, the prediction module is triggered to run; the prediction module is used to predict whether the framed and marked dynamic target is a risky dynamic target.

[0010] Furthermore, the surveillance video received by the annotation module is a spliced ​​video obtained by processing the surveillance images of several surveillance devices based on the panoramic splicing technology. During the operation phase of the annotation module, the dynamic targets are identified in the surveillance video based on the background difference method, and the identified dynamic targets are annotated with rectangular frames;

[0011] The labeling module is provided with a division unit and a storage unit at the lower level. The division unit is used to divide the monitoring video screen into a plurality of areas, and the plurality of areas correspond to the monitoring devices from which the areas in the monitoring video originate. The storage unit is used to intercept the images of the dynamic target areas marked by the running box of the labeling module in the monitoring video and store the images of the dynamic target areas.

[0012] Among them, the labeling module runs based on the specified operating frequency, and the division unit is manually operated by the system end user to divide the monitoring video screen, and the several divided areas are allowed to have partial overlapping areas with each other.

[0013] Furthermore, during the operation phase of the storage unit, when storing the dynamic target area images corresponding to the frame annotations, the storage unit simultaneously identifies the similarities between the dynamic target area images, and stores the dynamic target area images separately based on the similarity identification results;

[0014] The similarity calculation formula between the dynamic target area images is:

[0015] ;

[0016] Where: is the similarity between dynamic target area image a and dynamic target area image b; is the total amount of color levels in the dynamic target area image; is the number of pixels corresponding to the dynamic target area image a at the i-th color level; is the number of pixels corresponding to the dynamic target area image b at the i-th color level; is the calibration index; for The average value of the shortest distance from the center pixel to the edge of each corresponding dynamic target area image among the pixels corresponding to the maximum value of the ratio of the number of pixels corresponding to each color level within the range; for The average value of the shortest distance from the center pixel to the center pixel of each corresponding dynamic target area image among the pixels corresponding to the maximum ratio of the number of pixels corresponding to each color level within the range;

[0017] in, Express The averaging operation.

[0018] Furthermore, when the storage unit distinguishes and stores the dynamic target area images based on the similarity recognition result, a similarity distinction index is simultaneously set. Based on the similarity distinction index, the minimum similarity between the dynamic target area images in each differentiated storage interval is still greater than the similarity distinction index.

[0019] Furthermore, during the operation phase of the tracing module, an operation of selecting a dynamic target marked with a frame in the surveillance video, that is, an operation of selecting an image of a dynamic target region marked with a frame in any time frame in the surveillance video, after completing the selection operation, the selected dynamic target region image is used as a comparison target, and is paired with each differentiated storage interval in the storage unit, and the comprehensive similarity between the comparison target and each dynamic target region image in each paired differentiated storage interval is identified, and the differentiated storage interval with the highest comprehensive similarity is used as the object of the dynamic target tracing operation in the tracing module;

[0020] Among them, the comprehensive similarity stage of the dynamic target area image in the identification and comparison target and the distinguishing storage interval of each pair is based on The calculation logic is used to calculate the similarity between the recognition comparison target and the images of each dynamic target area in the distinguishing storage interval of each pair, and then the average of the similarity calculation results is obtained, and the obtained average is used as the comprehensive similarity between the comparison target and the images of each dynamic target area in the distinguishing storage interval of each pair.

[0021] Furthermore, after determining the dynamic target tracing operation object, the traceability module sorts the dynamic target area images in the differentiated storage interval pointed to by the dynamic target tracing operation object based on their source timestamps, and synchronously obtains the divided areas in the monitoring video screen to which each dynamic target area image belongs based on the sorting results, and then determines the source monitoring device of each dynamic target area image through the divided areas to complete the tracing of the dynamic target.

[0022] Furthermore, a segmentation unit is provided below the enhancement module and the ROI display module, and the segmentation unit is used to obtain the framed area for enhancement processing in the enhancement module, segment the framed area in the monitoring video, and send the segmented framed area video to the ROI display module;

[0023] The ROI display module is integrated with a mobile computer device with a display function. The ROI display module simultaneously displays several segmented framed area videos in a split-screen manner. When the ROI display module displays the segmented framed area videos, it adaptively scales the aspect ratio of the framed area videos so that the size of the framed area videos is consistent with the predetermined display area on the mobile computer.

[0024] Furthermore, the area selected by the enhancement module in the surveillance video includes the access control, gate, entrance and exit, fork in the road, and fire passage entrance and exit in the surveillance video, that is, the ROI area. The logic of the enhancement processing for the local enhancement target in the enhancement module is expressed as follows:

[0025] ;

[0026] Where: is the enhanced output pixel value; is the pixel value of the original video frame at position (x, y) and time t; is the ROI binary mask; is the detail enhancement coefficient; is the normalized distance; is the detail enhancement operator; is the noise suppression coefficient; is the noise estimate;

[0027] in, ; ; ;

[0028] Where: is the Euclidean distance from the pixel point (x, y) to the center of the ROI; is the equivalent radius of ROI; Indicates that the standard deviation is The Gaussian kernel is used to smooth the original image; represents the Laplace operator; Indicates that the standard deviation is The original image is smoothed by the Gaussian kernel, and the standard deviation of the Gaussian kernel is ∈[0.5, 1.5], Gaussian kernel standard deviation ∈[1.5, 3.0].

[0029] Furthermore, during the running phase of the prediction module, based on the tracing result of the frame-labeled dynamic target by the tracing module, the number of dynamic target area images in the storage interval corresponding to the frame-labeled dynamic target is obtained, which is recorded as q, and the number of independent display frame-selected areas in which the judgment module determines that the frame-labeled dynamic target exists is simultaneously obtained, which is recorded as m;

[0030] exist When a dynamic target marked with a box is predicted as a risky dynamic target, the differentiated storage interval in the storage unit corresponding to the risky dynamic target is locked simultaneously. The dynamic target area image in the locked differentiated storage interval cannot be replaced, modified, or deleted, and only new additions are supported;

[0031] Where: It is a custom judgment value;

[0032] in, The larger the value of , the more active the dynamic target annotated by the surface box is in the surveillance video area, but the lower the probability of spatial movement through access control, gates, entrances and exits, forks in the road, and fire passage entrances and exits. In other words, the larger the value, the more intentional the dynamic target annotated by the box is to avoid surveillance.

[0033] Furthermore, the labeling module is interactively connected to a division unit and a storage unit through a wireless network, the labeling module is interactively connected to a traceability module through a wireless network, the traceability module is interactively connected to the storage unit through a wireless network, the traceability module is interactively connected to an enhancement module and a ROI display module through a wireless network, the enhancement module and the ROI display module are interactively connected to a division unit through a wireless network, and the ROI display module is interactively connected to a judgment module and a prediction module through a wireless network.

[0034] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects:

[0035] The present invention provides an intelligent panoramic stitching video control system. During operation, the system can identify and mark dynamic targets based on the background difference method, divide the video screen to clarify the regional source device, and at the same time capture the dynamic target area image and store it separately according to the similarity. When the dynamic target is selected, it can be compared with the stored image, sorted by the source timestamp and the source device is determined in combination with the region to achieve traceability. Key areas such as access control can be framed for enhanced processing, and after segmentation, they can be displayed on a split screen on a mobile device. By counting the number of dynamic target images and the number of related display areas, it can be predicted whether it is a risk target, and the corresponding image interval of the risk target can be locked to ensure data security and improve the intelligence, accuracy and security of monitoring. This helps monitoring management personnel cope with daily monitoring and management of panoramic monitoring videos with a large amount of information. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0037] Figure 1 The figure is a structural diagram of an intelligent panoramic stitching video control system. DETAILED DESCRIPTION

[0038] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0039] The present invention will be further described below with reference to the embodiments.

[0040] Example:

[0041] An intelligent panoramic video splicing control system of this embodiment, such as Figure 1 As shown, including:

[0042] The annotation module is used to receive surveillance videos and annotate dynamic targets in the surveillance videos;

[0043] The surveillance video received by the annotation module is a spliced ​​video obtained by processing the surveillance images of several monitoring devices based on panoramic stitching technology. During the operation phase of the annotation module, dynamic targets are identified in the surveillance video based on the background difference method, and the identified dynamic targets are annotated with rectangular frames;

[0044] The labeling module is provided with a division unit and a storage unit at the lower level. The division unit is used to divide the monitoring video screen into a plurality of areas, and the plurality of areas correspond to the monitoring devices from which the areas in the monitoring video originate. The storage unit is used to intercept the images of the dynamic target areas marked by the running box of the labeling module in the monitoring video and store the images of the dynamic target areas.

[0045] The labeling module runs based on a specified operating frequency, and the division unit is manually operated by the system end user to divide the surveillance video screen, and the divided areas are allowed to have partial overlapping areas.

[0046] During the operation phase of the storage unit, when storing the dynamic target area images corresponding to the frame annotations, the similarities between the dynamic target area images are simultaneously identified, and the dynamic target area images are distinguished and stored based on the similarity identification results;

[0047] The formula for calculating the similarity between dynamic target area images is:

[0048] ;

[0049] Where: is the similarity between dynamic target area image a and dynamic target area image b; is the total amount of color levels in the dynamic target area image; is the number of pixels corresponding to the dynamic target area image a at the i-th color level; is the number of pixels corresponding to the dynamic target area image b at the i-th color level; is the calibration index; for The average value of the shortest distance from the center pixel to the edge of each corresponding dynamic target area image among the pixels corresponding to the maximum value of the ratio of the number of pixels corresponding to each color level within the range; for The average value of the shortest distance from the center pixel to the center pixel of each corresponding dynamic target area image among the pixels corresponding to the maximum ratio of the number of pixels corresponding to each color level within the range;

[0050] in, Express The mean operation of ;

[0051] It should be noted that the calibration index takes the value of 1 or -1. When the numerator of the fraction containing the calibration index is less than or equal to the denominator, the calibration index takes the value of 1. Otherwise, the calibration index takes the value of -1.

[0052] Through the above logic formula calculation, a formula for calculating the similarity between dynamic target area images is provided, which provides necessary operation data support for the subsequent module operation of the system in this embodiment;

[0053] When the storage unit distinguishes and stores the dynamic target area images based on the similarity recognition result, a similarity distinction index is simultaneously set, and based on the similarity distinction index, the minimum similarity value between the dynamic target area images in each distinguished storage interval is still greater than the similarity distinction index;

[0054] The tracing module is used to select dynamic targets marked in the frames in the surveillance video and trace the dynamic targets;

[0055] During the tracing module operation phase, the operation of selecting the dynamic target marked by the frame in the surveillance video, that is, the operation of selecting the image of the dynamic target area marked by the frame in any time frame in the surveillance video, after completing the selection operation, the selected dynamic target area image is used as the comparison target, and paired with each differentiated storage interval in the storage unit respectively, and the comprehensive similarity between the comparison target and each dynamic target area image in each paired differentiated storage interval is identified, and the differentiated storage interval with the highest comprehensive similarity is used as the object of the dynamic target tracing operation in the tracing module;

[0056] Among them, the comprehensive similarity stage of the dynamic target area image in the identification and comparison target and the distinguishing storage interval of each pair is based on The calculation logic is used to calculate the similarity between the recognition comparison target and the images of each dynamic target region in each paired distinguishing storage interval, and then the average of the similarity calculation results is obtained, and the average is used as the comprehensive similarity between the comparison target and the images of each dynamic target region in each paired distinguishing storage interval;

[0057] After determining the dynamic target tracing operation object, the traceability module sorts the dynamic target area images in the storage interval pointed to by the dynamic target tracing operation object based on their source timestamps, and synchronously obtains the divided areas in the monitoring video screen to which each dynamic target area image belongs based on the sorting results. Then, the source monitoring device of each dynamic target area image is determined by dividing the areas to complete the traceability of the dynamic target;

[0058] The enhancement module is used to select an area in the surveillance video as a local enhancement target and perform enhancement processing on the local enhancement target;

[0059] The area selected by the enhancement module in the surveillance video includes access control, gates, entrances and exits, forks in the road, and fire passage entrances and exits in the surveillance video, namely the ROI area. The logic of the enhancement processing for the local enhancement target in the enhancement module is expressed as follows:

[0060] ;

[0061] Where: is the enhanced output pixel value; is the pixel value of the original video frame at position (x, y) and time t; is the ROI binary mask; is the detail enhancement coefficient; is the normalized distance; is the detail enhancement operator; is the noise suppression coefficient; is the noise estimate;

[0062] in, ; ; ;

[0063] Where: is the Euclidean distance from the pixel point (x, y) to the center of the ROI; is the equivalent radius of ROI; Indicates that the standard deviation is The Gaussian kernel is used to smooth the original image; represents the Laplace operator; Indicates that the standard deviation is The original image is smoothed by the Gaussian kernel, and the standard deviation of the Gaussian kernel is ∈[0.5, 1.5], Gaussian kernel standard deviation ∈[1.5, 3.0];

[0064] The above logic formula further defines the enhancement logic of the video corresponding to the selected area in the surveillance video as the local enhancement target.

[0065] The ROI display module is used to receive the framed area enhanced by the enhancement module and independently display the framed area;

[0066] A segmentation unit is provided below the enhancement module and the ROI display module. The segmentation unit is used to obtain the framed area for enhancement processing in the enhancement module, segment the framed area in the monitoring video, and send the segmented framed area video to the ROI display module;

[0067] The ROI display module is integrated with a mobile computer device having a display function. The ROI display module displays several segmented framed area videos simultaneously in a split-screen manner. When displaying the segmented framed area videos, the ROI display module adaptively scales the aspect ratio of the framed area videos so that the size of the framed area videos is consistent with the predetermined display area on the mobile computer.

[0068] The judgment module is used to judge whether there is a dynamic target marked by the box selected by the traceability module in the historical video of the frame selection area independently displayed in the ROI display module. If the judgment result is yes, the prediction module is triggered to run;

[0069] Prediction module, used to predict whether the dynamic target marked by the box is a risky dynamic target;

[0070] During the prediction module's operation phase, based on the traceability results of the frame-labeled dynamic targets by the traceability module, the number of dynamic target area images in the storage interval corresponding to the frame-labeled dynamic targets is obtained, denoted as q. Simultaneously, the number of independently displayed frame-selected areas where the judgment module determines that a frame-labeled dynamic target exists is obtained, denoted as m.

[0071] exist When a dynamic target marked with a box is predicted as a risky dynamic target, the differentiated storage interval in the storage unit corresponding to the risky dynamic target is locked simultaneously. The dynamic target area image in the locked differentiated storage interval cannot be replaced, modified, or deleted, and only new additions are supported;

[0072] Where: It is a custom judgment value;

[0073] in, The larger the value of , the more active the dynamic target marked by the surface box is in the surveillance video area, but the lower the probability of spatial movement through access control, gates, entrances and exits, forks in the road, and fire passage entrances and exits. In other words, the larger the value, the more intentional the dynamic target marked by the box is to avoid surveillance.

[0074] The labeling module is interactively connected to the division unit and the storage unit through a wireless network, the labeling module is interactively connected to the traceability module through a wireless network, the traceability module is interactively connected to the storage unit through a wireless network, the traceability module is interactively connected to the enhancement module and the ROI display module through a wireless network, the enhancement module and the ROI display module are interactively connected to the division unit through a wireless network, and the ROI display module is interactively connected to the judgment module and the prediction module through a wireless network.

[0075] In this embodiment, the labeling module receives the surveillance video and frames the dynamic targets in the surveillance video. The segmentation unit simultaneously divides the surveillance video to obtain a plurality of regions, each of which corresponds to the source surveillance device of each region in the surveillance video. The storage unit captures the images of the dynamic target regions framed and marked by the labeling module in the surveillance video in real time and stores the images of the dynamic target regions. The tracing module selects the dynamic targets framed and marked in the surveillance video and traces the dynamic targets. The enhancement module then frames an area in the surveillance video as a local enhanced target and enhances the local enhanced target. The ROI display module further receives the framed area enhanced by the enhancement module and independently displays the framed area. The segmentation unit simultaneously obtains the framed area enhanced by the enhancement module, segments the framed area in the surveillance video, and sends the segmented framed area video to the ROI display module. The determination module determines whether the framed and marked dynamic target selected by the tracing module is present in the historical video of the framed area independently displayed in the ROI display module. If the determination result is yes, the prediction module is triggered to run. Finally, the prediction module predicts whether the framed and marked dynamic target is a risky dynamic target.

[0076] Through the operation of the system in the above embodiment, the processing of panoramic surveillance video can be handled so that the user end can more quickly grasp the key information in the panoramic surveillance video and further optimize the information display performance of the panoramic surveillance video in actual application scenarios.

[0077] Based on the above embodiment, the system in the above embodiment is applied below to provide a brief example of an extensible system application:

[0078] xx large commercial complex (including shopping malls, office buildings, and underground parking lots) deploys an intelligent panoramic stitching video control system. The system accesses images from multiple monitoring devices and generates stitching videos covering the entire scene through panoramic stitching technology, focusing on monitoring areas such as office building lobby entrances and exits, underground parking lot vehicle entrances and exits, escalator entrances on each floor of the shopping mall, fire escape entrances and exits, and fork in the road.

[0079] 2. System Operation Process

[0080] (1) Dynamic target labeling and storage

[0081] The labeling module runs at a fixed frequency, using background subtraction to identify dynamic objects (such as pedestrians and vehicles) in the spliced ​​video and annotate them with rectangular boxes. For example, if a pedestrian in a red coat is detected moving in the first-floor atrium, or a black car is detected at the entrance of the underground parking lot, the system will capture and store the image of the area.

[0082] Security personnel manually divide the surveillance screen into 20 areas, each area corresponding to specific surveillance equipment (such as office building lobbies, shopping mall floors, etc.), and partial overlap is allowed between areas.

[0083] The storage unit performs similarity calculation and classified storage on dynamic target images: images with similar color and pixel distribution (such as pedestrians wearing the same color coat) are stored in the same interval, and images with large differences (such as vehicles of different models) are stored separately.

[0084] (2) Dynamic target tracing

[0085] The security personnel selects a dynamic target (such as an image of a pedestrian wearing a red jacket) in the monitoring image. The system uses this image as the comparison target and calculates the comprehensive similarity (based on features such as color, pixel distribution, and spatial position) with the images in each storage interval to match the interval with the highest similarity.

[0086] The system analyzes the target's movement trajectory based on the chronological order of images within that interval. For example, a pedestrian in a red jacket appears in the office building lobby at 9:00 AM, at the first-floor escalator entrance at 9:15 AM, and at the second-floor fire exit at 9:30 AM. By combining the corresponding surveillance equipment in the area, the system can determine the target's movement path.

[0087] (3) Local area enhancement and display

[0088] The enhancement module enhances the image quality of key areas (such as the entrance and exit of the fire escape on the second floor), highlighting facial details of people, vehicle license plates and other features, and reducing noise interference.

[0089] The system separates the enhanced key area video from the original image and sends it to the security personnel's mobile tablet. The tablet displays multiple key area videos simultaneously in split-screen format, automatically adjusting the aspect ratio to fit the screen.

[0090] (IV) Risk target prediction

[0091] The system compares the video displayed in key areas with historical images and counts the number of times a certain target appears in the key area (for example, a pedestrian in a red jacket appears three times at the entrance and exit of the fire escape on the second floor).

[0092] Based on the target's total number of active events (e.g., 15 appearances), the system calculates the ratio of its activity level to its movement frequency in key areas. If this ratio exceeds a preset threshold, the system identifies the target as a "risky dynamic target" because it is highly active but rarely passes through key areas like access control points and entrances, suggesting it is deliberately avoiding primary monitoring routes.

[0093] The system automatically locks the storage area of ​​the target. Images in this area can only be added but cannot be modified or deleted. At the same time, an early warning is sent to the security center.

[0094] In summary, in the above embodiment, during operation, the system can identify and mark dynamic targets based on the background difference method, divide the video screen to clarify the regional source device, and at the same time capture the dynamic target area image and store it separately according to the similarity. After selecting the dynamic target, it can be compared with the stored image, sorted by the source timestamp and combined with the region to determine the source device to achieve traceability. For key areas such as access control, frame selection and enhancement processing can be performed, and after segmentation, it can be displayed on a split screen on a mobile device. By counting the number of dynamic target images and the number of related display areas, it can be predicted whether it is a risk target, and the corresponding image interval of the risk target can be locked to ensure data security and improve the intelligence, accuracy and security of monitoring, so as to cooperate with monitoring management personnel to cope with the daily monitoring and management of panoramic surveillance videos with huge amounts of information.

[0095] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An intelligent panoramic video stitching control system, characterized in that: include: The annotation module is used to receive surveillance videos and annotate dynamic targets in the surveillance videos; The labeling module is provided with a division unit and a storage unit at the lower level. The division unit is used to divide the monitoring video screen into a plurality of areas, and the plurality of areas correspond to the monitoring devices from which the areas in the monitoring video originate. The storage unit is used to intercept the images of the dynamic target areas marked by the labeling module frame in the monitoring video and store the images of the dynamic target areas. During the operation phase of the storage unit, when storing the dynamic target area images corresponding to the frame marks, the storage unit simultaneously identifies the similarities between the dynamic target area images, and stores the dynamic target area images separately based on the similarity identification results; The similarity calculation formula between the dynamic target area images is: ； Where: is the similarity between dynamic target area image a and dynamic target area image b; is the total amount of color levels in the dynamic target area image; is the number of pixels corresponding to the dynamic target area image a at the i-th color level; is the number of pixels corresponding to the dynamic target area image b at the i-th color level; is the calibration index; for The average value of the shortest distance from the center pixel to the edge of each corresponding dynamic target area image among the pixels corresponding to the maximum value of the ratio of the number of pixels corresponding to each color level within the range; for The average value of the shortest distance from the center pixel to the center pixel of each corresponding dynamic target area image among the pixels corresponding to the maximum ratio of the number of pixels corresponding to each color level within the range; in, Express The mean operation of ; The tracing module is used to select dynamic targets marked in the frames in the surveillance video and trace the dynamic targets; The enhancement module is used to select an area in the surveillance video as a local enhancement target and perform enhancement processing on the local enhancement target; The ROI display module is used to receive the framed area enhanced by the enhancement module and independently display the framed area; The judgment module is used to judge whether there is a dynamic target marked by the box selected by the traceability module in the historical video of the frame selection area independently displayed in the ROI display module. If the judgment result is yes, the prediction module is triggered to run; Prediction module, used to predict whether the dynamic target marked by the box is a risky dynamic target; During the running phase of the prediction module, based on the tracing result of the frame-labeled dynamic target by the tracing module, the number of dynamic target area images in the storage interval corresponding to the frame-labeled dynamic target is obtained, which is recorded as q. At the same time, the number of independent display frame-selected areas determined by the determination module to contain the frame-labeled dynamic target is obtained, which is recorded as m. exist When a dynamic target marked with a box is predicted as a risky dynamic target, the differentiated storage interval in the storage unit corresponding to the risky dynamic target is locked simultaneously. The dynamic target area image in the locked differentiated storage interval cannot be replaced, modified, or deleted, and only new additions are supported; Where: It is a custom judgment value; in, The larger the value of , the more active the dynamic target marked in the frame is in the surveillance video area, but the lower the probability of spatial movement through access control, gates, entrances and exits, forks in the road, and fire passage entrances and exits. In other words, the larger the value, the more intentional the dynamic target marked in the frame is to avoid surveillance.

2. The intelligent panoramic video stitching control system according to claim 1, characterized in that: The surveillance video received by the annotation module is a spliced ​​video obtained by processing the surveillance images of several monitoring devices based on the panoramic splicing technology. During the operation phase of the annotation module, dynamic targets are identified in the surveillance video based on the background difference method, and the identified dynamic targets are annotated with rectangular frames; Among them, the labeling module runs based on the specified operating frequency, and the division unit is manually operated by the system end user to divide the monitoring video screen, and the several divided areas are allowed to have partial overlapping areas with each other.

3. The intelligent panoramic video stitching control system according to claim 1, characterized in that: When the storage unit distinguishes and stores the dynamic target area images based on the similarity recognition result, a similarity distinction index is simultaneously set. Based on the similarity distinction index, the minimum similarity between the dynamic target area images in each distinguished storage interval is still greater than the similarity distinction index.

4. The intelligent panoramic video stitching control system according to claim 1, characterized in that: During the operation phase of the tracing module, an operation is performed to select a dynamic target marked in a frame in the surveillance video, that is, an operation is performed to select a frame marked corresponding to a dynamic target area image in any frame in the surveillance video. After the selection operation is completed, the selected dynamic target area image is used as a comparison target, and paired with each differentiated storage interval in the storage unit, and the comprehensive similarity between the comparison target and each dynamic target area image in each paired differentiated storage interval is identified. The differentiated storage interval with the highest comprehensive similarity is used as the object of the dynamic target tracing operation in the tracing module; Among them, the comprehensive similarity stage of the dynamic target area image in the identification and comparison target and the distinguishing storage interval of each pair is based on The calculation logic is used to calculate the similarity between the recognition comparison target and the images of each dynamic target area in the distinguishing storage interval of each pair, and then the average of the similarity calculation results is obtained, and the obtained average is used as the comprehensive similarity between the comparison target and the images of each dynamic target area in the distinguishing storage interval of each pair.

5. The intelligent panoramic video stitching control system according to claim 1, characterized in that: After determining the dynamic target tracing operation object, the tracing module sorts the dynamic target area images in the differentiated storage interval pointed to by the dynamic target tracing operation object based on their source timestamps, and synchronously obtains the divided areas in the monitoring video screen to which each dynamic target area image belongs based on the sorting results, and then determines the source monitoring device of each dynamic target area image through the divided areas to complete the tracing of the dynamic target.

6. The intelligent panoramic video stitching control system according to claim 1, characterized in that: The enhancement module and the ROI display module are provided with a segmentation unit at the lower level, which is used to obtain the framed area for enhancement processing in the enhancement module, segment the framed area in the monitoring video, and send the segmented framed area video to the ROI display module; The ROI display module is integrated with a mobile computer device with a display function. The ROI display module simultaneously displays several segmented framed area videos in a split-screen manner. When the ROI display module displays the segmented framed area videos, it adaptively scales the aspect ratio of the framed area videos so that the size of the framed area videos is consistent with the predetermined display area on the mobile computer.

7. The intelligent panoramic video stitching control system according to claim 1, characterized in that: The area selected by the enhancement module in the surveillance video includes the access control, gate, entrance and exit, fork in the road, and entrance and exit of the fire passage in the surveillance video, that is, the ROI area. The logic of the enhancement processing for the local enhancement target in the enhancement module is expressed as follows: ； Where: is the enhanced output pixel value; is the pixel value of the original video frame at position (x, y) and time t; is the ROI binary mask; is the detail enhancement coefficient; is the normalized distance; is the detail enhancement operator; is the noise suppression coefficient; is the noise estimate; in, ; ; ; Where: is the Euclidean distance from the pixel point (x, y) to the center of the ROI; is the equivalent radius of ROI; Indicates that the standard deviation is The Gaussian kernel is used to smooth the original image; represents the Laplace operator; Indicates that the standard deviation is The original image is smoothed by the Gaussian kernel, and the standard deviation of the Gaussian kernel is ∈[0.5, 1.5], Gaussian kernel standard deviation ∈[1.5, 3.0].

8. The intelligent panoramic video stitching control system according to claim 1, characterized in that: The labeling module is interactively connected to the division unit and the storage unit through a wireless network, the labeling module is interactively connected to the traceability module through a wireless network, the traceability module is interactively connected to the storage unit through a wireless network, the traceability module is interactively connected to the enhancement module and the ROI display module through a wireless network, the enhancement module and the ROI display module are interactively connected to the division unit through a wireless network, and the ROI display module is interactively connected to the judgment module and the prediction module through a wireless network.

Images