Video monitoring processing method and system
By filtering and processing video surveillance data, using target characteristics and behavioral logic to form an advanced surveillance video library, and 3D adjustments are made, the problems of manual viewing and low quality of surveillance videos in the existing technology are solved, and more efficient and accurate video surveillance processing.
Patent Information
- Application Number
- CN202410838696.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-26
- Publication Date
- 2024-12-06
AI Technical Summary
The existing video surveillance system relies on manual viewing, which leads to huge resource consumption and is prone to error and missed inspections. The unprocessed surveillance video quality is low and it is difficult to track the target.
By obtaining the characteristics of the tracking target and the latest information to appear, filter the monitoring videos that meet the regional requirements, filter according to the target characteristics and behavioral logic, form an advanced monitoring video library, and use three-dimensional spatial diagrams to adjust the video picture, filter important information and push it to User side.
It improves the convenience and accuracy of video surveillance processing, reduces unnecessary video interference, enhances image quality, and improves processing speed and efficiency.
Smart Images

Figure CN119094695A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of video processing, and in particular to a video surveillance processing method and system. Background Art
[0002] Video surveillance is a technology and system that uses cameras and related equipment to monitor and record activities in a specific area or place in real time. Video surveillance is a core component of the construction of a "safe city" and an important guarantee for urban public safety.
[0003] In the related art, tracking targets is still done by manual review, and the number of surveillance videos is huge. Manual review to find targets consumes a lot of human resources. In addition, the unprocessed surveillance videos have poor video effects and are prone to mis-checking and missed inspections. The processing of video surveillance is very inconvenient and needs improvement. Summary of the invention
[0004] The purpose of the present invention is to provide a video monitoring processing method and system to solve the problems raised in the above background technology.
[0005] In the first aspect, the present application provides a video surveillance processing method, which adopts the following technical solution: Obtain at least one feature of the tracking target and record it as the target feature, extract the most recent appearance information of the tracking target and record it as the last information; Obtain monitoring data from video surveillance, and select monitoring videos that meet regional requirements from the monitoring data based on the latest information to form a primary monitoring video library; Based on the target feature, an information volume threshold is set, and surveillance videos reaching the information volume threshold are selected from the primary surveillance video library to form an intermediate surveillance video library; Extract the tracking target behavior of the intermediate surveillance video, and filter the surveillance videos that meet the logic and interaction conditions according to the tracking target behavior to form an advanced surveillance video library; Based on the advanced surveillance video library, a target position of the tracking target is obtained, a three-dimensional spatial map of the target feature is formed according to the target position, and the advanced surveillance video picture is adjusted according to the three-dimensional spatial map to obtain a 3D surveillance video library; Obtain the importance of information in different areas of the surveillance video, filter it according to the importance of the information, obtain valid video and push it to the user end.
[0006] Preferably, the step of obtaining at least one feature of the tracking target and recording it as the target feature, and extracting the most recent appearance information of the tracking target and recording it as the last occurrence information, specifically comprises: Obtain at least one feature of the tracking target and record it as the target feature, determine whether the target feature is a fixed feature, and if the target feature is a fixed feature, classify the target feature as a main feature; If the target feature is not a fixed feature, the target feature is classified as an auxiliary feature; Get the category to which the tracking target belongs, and form a corresponding level according to the classification unit of the category; Match the classification unit to which the main feature belongs, find the level corresponding to the classification unit, and obtain the feature level of the main feature; The last information includes the last location and the last time point.
[0007] Preferably, the step of obtaining the monitoring data of the video surveillance and selecting the monitoring videos meeting the regional requirements from the monitoring data according to the last information to form a primary monitoring video library is specifically as follows: Get the real-time point, and calculate the time difference between the real-time point and the appearance time point; Obtain a traffic operation route map, and determine the maximum moving distance of the tracking target based on the traffic operation route map and the time difference; With the last location as the center and the maximum moving distance as the radius, a search area is formed; Extract the monitoring locations of all monitoring videos from the monitoring data and determine whether the monitoring locations are within the search area; If the monitoring location is within the search area, the corresponding monitoring video will form a primary monitoring video library.
[0008] Preferably, based on the target feature, an information volume threshold is set, and the step of selecting surveillance videos reaching the information volume threshold from the primary surveillance video library to form an intermediate surveillance video library is specifically as follows: Compare the target features with the primary surveillance video to obtain the video similarity AP; Extract the duration BS of the target being tracked in the primary surveillance video, and extract the ratio BB of the area of the target being tracked in the primary surveillance video to the area of the target being tracked; According to the information related function The information volume CX is calculated, where , , is the scale factor and is greater than 0; An information volume threshold is set, and surveillance videos that reach the information volume threshold are selected from the primary surveillance video library to form an intermediate surveillance video library.
[0009] Preferably, the step of comparing the target feature with the primary surveillance video to obtain the video similarity AP is specifically as follows: Set the feature weight ratio of different main features according to the feature level , where i is the number of the corresponding level of different main characteristics; Scan the primary surveillance video, compare it according to the target features, and obtain the suspected target in the primary surveillance video; Extract the suspected features of the suspected target, compare the suspected features with the main features one by one, and obtain the feature similarity between the suspected features and the main features; Set a feature similarity threshold, and select suspected features whose feature similarity reaches the feature similarity threshold as the first feature; Find the feature level corresponding to the first feature and count the number of first features at different feature levels And the corresponding feature similarity is , where i is the number of the levels corresponding to different main features, and j is the feature similarity number corresponding to the first feature; Compare the suspected features with the auxiliary features one by one, and count the number AL of suspected features whose feature similarity reaches the feature similarity threshold; According to the similarity correlation function Calculate the video similarity AP, where n is The maximum number of m, the maximum number of levels corresponding to different main features, is , is the scaling factor and is greater than 0.
[0010] Preferably, the step of scanning the primary surveillance video and comparing according to the target features to obtain the suspected target in the primary surveillance video is specifically: Scanning the primary surveillance video to extract video information in the primary surveillance video, wherein the video information includes video brightness, video weather, and video background color; Acquire standard information of target features, wherein the standard information includes standard brightness, standard weather, and standard background color; Compare the video information and the standard information to see if they are consistent. If they are inconsistent, adjust the target features according to the changes in the video information to obtain the comparison features. If the video information is consistent with the standard information, the target feature is directly used as the comparison feature; Search for targets with similar features to the target in the primary surveillance video as suspected targets.
[0011] Preferably, the step of extracting the tracking target behavior of the intermediate monitoring video and screening the monitoring videos that meet the logic and interaction conditions according to the tracking target behavior to form the advanced monitoring video library is specifically as follows: Extract the moving trajectory of the target in the intermediate surveillance video and determine whether the moving trajectory follows the laws of physics; If the movement trajectory does not follow the laws of physics, the corresponding intermediate surveillance video is judged to be unqualified; Obtain the interaction between the tracking target and other elements in the intermediate surveillance video, and determine whether the interaction conforms to scientific common sense. If the interaction does not conform to scientific common sense, the corresponding intermediate surveillance video is determined to be unqualified. Intermediate surveillance video libraries that do not meet the requirements are screened and removed to obtain advanced surveillance video libraries.
[0012] Preferably, based on the advanced surveillance video library, the target position of the tracking target is obtained, a three-dimensional spatial map of the target feature is formed according to the target position, and the advanced surveillance video picture is adjusted according to the three-dimensional spatial map to obtain the 3D surveillance video library, specifically, the steps are: Obtain the monitoring range of all advanced monitoring videos and extract the target position of the tracking target within the monitoring range; Advanced surveillance videos with the same target location are grouped into one category and recorded as same-location surveillance videos; Extract the monitoring picture of the tracking target at the same time in the monitoring video of the same location to form a three-dimensional spatial map of the tracking target; The relative position of the tracking target and other objects in the surveillance video is obtained according to the three-dimensional space diagram, and the spatial position of the tracking target is reflected by adjusting the relative size of the tracking target and other objects; The order of the distances of the tracking target and other objects from the monitoring device is counted, and the color brightness of the tracking target and other objects is increased from far to near according to the order of distances, and a 3D monitoring video library is obtained.
[0013] Preferably, the steps of obtaining the importance of information in different areas of the surveillance video, filtering according to the importance of the information, obtaining valid videos and pushing them to the user end are specifically as follows: Divide the area according to the objects in the surveillance video to obtain different object areas; Extract the object area features to find similar surveillance videos in the surveillance data and count the number of similar surveillance videos DS; Based on the interaction situation, extract the interaction times DC and interaction times DJ between different objects and the tracking target; According to the importance of information Calculate the information importance DZ, where , , is the scale factor and is greater than 0; The information importance of the suspected target area is the highest value. The corresponding filtering degree is set according to the information importance. After filtering different object areas, valid videos are obtained and pushed to the user end.
[0014] In a second aspect, the present application provides a video surveillance processing system, which adopts the following technical solution: A video surveillance processing system, a target feature module, obtains at least one feature of a tracking target and records it as a target feature, extracts the most recent appearance information of the tracking target and records it as the last occurrence information; The primary screening module obtains the monitoring data of the video surveillance, and screens the monitoring videos that meet the regional requirements from the monitoring data according to the last information to form a primary monitoring video library; The intermediate screening module sets an information volume threshold based on the target feature, and selects surveillance videos that reach the information volume threshold from the primary surveillance video library to form an intermediate surveillance video library; The advanced screening module extracts the tracking target behavior of the intermediate surveillance video, and screens the surveillance videos that meet the logic and interaction conditions according to the tracking target behavior to form an advanced surveillance video library; A 3D adjustment module, based on the advanced surveillance video library, obtains a target position of a tracking target, forms a three-dimensional spatial map of target features according to the target position, and adjusts the advanced surveillance video picture according to the three-dimensional spatial map to obtain a 3D surveillance video library; The filtering and pushing module obtains the importance of information in different areas of the surveillance video, filters it according to the importance of the information, obtains valid video and pushes it to the user end.
[0015] In summary, the present application includes at least one of the following beneficial technical effects: 1. The surveillance videos of the areas that the tracking target cannot reach are filtered out by the last appearance information of the tracking target. Then, the surveillance videos that do not meet the target characteristics are filtered out according to the feature comparison of the tracking target. The surveillance videos that do not meet the objective laws are filtered out according to the logic and interaction of the target behavior. Finally, the remaining surveillance videos are processed and filtered with 3D effects and pushed to the user end, which is conducive to obtaining the target surveillance video quickly and conveniently, reducing unnecessary video interference, and improving the convenience and accuracy of video surveillance processing.
[0016] 2. The features of the tracking target are distinguished according to whether the features are fixed, and divided into main features and auxiliary features. The main features are graded according to the classification units of the categories to which they belong, so as to calculate the similarity between different surveillance videos and the tracking target. Combined with the duration of the tracking target and the area occupied, the amount of information in the surveillance video is judged, and the effective video is screened according to the set information threshold. Not only invalid videos without the tracking target are removed, but also videos that contain the tracking target but do not play any role are excluded, so as to obtain more accurate and effective surveillance videos and improve the accuracy of video surveillance processing.
[0017] 3. When filtering the surveillance video, the surveillance area is divided into different areas according to different objects, and the number of similar surveillance videos in different areas is found. At the same time, the importance of information in different areas is determined based on the interaction time and number of interactions between the object and the tracking target, thereby determining the degree of filtering. On the basis of improving image quality, the filtering speed is accelerated, the resources required for filtering are reduced, and the speed of video surveillance processing is improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 It is a schematic diagram of specific steps of an embodiment of a video monitoring processing method of the present invention.
[0019] Figure 2 It is a module connection diagram of an embodiment of a video monitoring processing system of the present invention. DETAILED DESCRIPTION
[0020] Below is a combination of the embodiments and Figure 1-Figure 2 The present invention is further described in detail, but the embodiments of the present invention are not limited thereto.
[0021] The present invention discloses a video monitoring processing method, which specifically comprises the following steps: Step S1, obtaining at least one feature of the tracking target and recording it as the target feature, extracting the most recent appearance information of the tracking target and recording it as the last occurrence information.
[0022] Step S2, obtaining monitoring data of video surveillance, and screening out monitoring videos that meet regional requirements from the monitoring data according to the latest information to form a primary monitoring video library.
[0023] Step S3, based on the target features, an information volume threshold is set, and surveillance videos that reach the information volume threshold are selected from the primary surveillance video library to form an intermediate surveillance video library.
[0024] Step S4, extracting the tracking target behavior of the intermediate monitoring video, and screening the monitoring videos that meet the logic and interaction conditions according to the tracking target behavior to form an advanced monitoring video library.
[0025] Step S5, based on the advanced monitoring video library, obtain the target position of the tracking target, form a three-dimensional space map of the target characteristics according to the target position, adjust the advanced monitoring video picture according to the three-dimensional space map, and obtain a 3D monitoring video library.
[0026] Step S6, obtaining the importance of information in different areas of the surveillance video, filtering according to the importance of the information, obtaining valid videos and pushing them to the user end.
[0027] In actual application, the tracking target is still done manually, and the number of surveillance videos is huge. It will consume a lot of human resources to search for targets manually. In addition, a large number of video resources are not processed, and it is difficult to identify the tracking target due to their low pixels. At the same time, some tracking targets, such as puppies, are similar and difficult to distinguish with the naked eye. By tracking the last information of the target, the amount of surveillance video information, and the logic and interaction of the target behavior, the relevant surveillance video is pushed to the user end after removing invalid videos, which can greatly improve the tracking efficiency. At the same time, filtering the surveillance video and enhancing the 3D effect of the video can help to obtain accurate tracking target information more clearly and quickly.
[0028] The steps of obtaining at least one feature of the tracking target and recording it as the target feature, and extracting the most recent appearance information of the tracking target and recording it as the last appearance information are specifically as follows: Step S11, obtaining at least one feature of the tracking target and recording it as the target feature, determining whether the target feature is a fixed feature, and if the target feature is a fixed feature, classifying the target feature as a main feature.
[0029] Step S12: if the target feature is not a fixed feature, the target feature is classified as an auxiliary feature.
[0030] Step S13, obtaining the category to which the tracking target belongs, and forming a corresponding level according to the classification unit of the category.
[0031] Step S14, matching the classification unit to which the main feature belongs, searching for the level corresponding to the classification unit, and obtaining the feature level of the main feature.
[0032] The larger the classification unit, the lower the corresponding feature level. The higher the feature level, the greater the role of the feature in tracking the target.
[0033] Step S15, the last information includes the last location and the last time point.
[0034] In actual application, a tracking target has many features, and it is necessary to distinguish the target features, which is conducive to the subsequent judgment of whether it is a tracking target. Fixed features are more accurate in matching objects, while non-fixed features can only play an auxiliary role. For example, when tracking a target, the target's appearance, height, etc. are fixed features and difficult to change. The target's clothes and dress are non-fixed features and can be changed at any time, so they can only play an auxiliary role. The smaller the classification unit of the category described in the fixed feature, the more similar the organisms in the unit are, the more common features there are, the smaller the difference between the organisms, the fewer species of organisms are included, and the more accurate the comparison result. For example, the tracking target is a Samoyed dog, and the features of "pointed head and palate, long face, protruding nose, pointed and upright ears" belong to the canine features. The classification unit is "family" and the classification unit is large. Therefore, it is easy to find many canine animals by searching for targets based on this feature. "Scars on the legs" is the feature that distinguishes this Samoyed dog from other Samoyed dogs. Therefore, the classification unit of "scars on the legs" is very small, and the results obtained by searching based on this feature will be more accurate.
[0035] The steps of obtaining the monitoring data of the video surveillance and selecting the monitoring videos that meet the regional requirements from the monitoring data according to the last information to form a primary monitoring video library are as follows: Step S21, obtaining the real-time point, and calculating the time difference between the real-time point and the appearance time point.
[0036] Step S22, obtaining a traffic operation route map, and determining the maximum moving distance of the tracking target according to the traffic operation route map and the time difference.
[0037] Step S23, forming a search area with the last location as the center and the maximum moving distance as the radius.
[0038] Step S24, extracting the monitoring locations of all monitoring videos from the monitoring data, and determining whether the monitoring locations are within the search area.
[0039] Step S25: If the monitoring location is within the search area, the corresponding monitoring video is formed into a primary monitoring video library.
[0040] In actual use, there are countless surveillance cameras across the country. Screening all surveillance cameras will increase the workload of screening and increase the tracking time. According to the last information, some impossible areas can be excluded, thereby narrowing the monitoring scope. For example, if we want to track target A now, the last time target A appeared was at point A in Beijing at 14:00, and the current time is 14:30. It takes one hour to reach the airport and half an hour to reach the high-speed rail station. No speeding was found in Beijing. The maximum speed limit for cars is 120 kilometers per hour. Excluding airplanes and high-speed rail, the fastest means of transportation available to the current target is cars. Therefore, a circular area with a radius of 120 kilometers is formed with point A as the center, and all surveillance videos in the area are formed into primary surveillance videos. Many unnecessary videos can be excluded, thereby improving the processing speed. If the target arrives at the high-speed rail station within half an hour, it is necessary to set the radius in conjunction with the nearest high-speed rail at the high-speed rail station and the speed of the high-speed rail. The same is true for other means of transportation such as airplanes.
[0041] Based on the target features, the information volume threshold is set, and the steps of selecting surveillance videos that reach the information volume threshold from the primary surveillance video library to form an intermediate surveillance video library are as follows: Step S31, compare the target feature with the primary surveillance video to obtain the video similarity AP.
[0042] Step S32, extracting the duration BS of the tracking target appearing in the primary monitoring video, and extracting the ratio BB of the screen area of the tracking target appearing in the primary monitoring video to the tracking target area.
[0043] Step S33, according to the information volume correlation function The information volume CX is calculated, where , , is the scaling factor and is greater than 0.
[0044] Step S34, setting an information volume threshold, and selecting surveillance videos that reach the information volume threshold from the primary surveillance video library to form an intermediate surveillance video library.
[0045] In actual use, in the search area, the tracking target will only pass through some areas, so many surveillance videos will not capture the tracking target, which is also an invalid surveillance video. First, compare according to the target characteristics and filter out the surveillance videos where the tracking target does not appear. Secondly, if the tracking target appears for too short a time, or the tracking target appears in a small area in the surveillance video, then the role of the surveillance screen is not very large. Therefore, combined with the video similarity and the actual monitoring situation, confirm whether to filter out the surveillance. While deleting the surveillance video, reduce the loss of effective videos. For example, the tracking target similarity in surveillance video A is low, and the area ratio of the tracking target is only 30%, and the appearance time is 5 seconds. Therefore, the surveillance video provides less effective information. On the one hand, it cannot be fully confirmed whether it is the tracking target. On the other hand, the screen area is small and the time is short, and it is also impossible to confirm whether it is the tracking target. At the same time, the behavior of the tracking target is not obtained, and it is difficult to confirm the next action, so the video can be filtered out.
[0046] The steps of comparing the target features with the primary surveillance video to obtain the video similarity AP are as follows: Step S311, setting feature weight ratios of different main features according to feature levels , where i is the number of the corresponding level of different main characteristics.
[0047] Step S312, scanning the primary surveillance video, comparing it according to the target features, and obtaining the suspected target in the primary surveillance video.
[0048] Step S313, extracting suspected features of the suspected target, comparing the suspected features with the main features one by one, and obtaining feature similarity between the suspected features and the main features.
[0049] Step S314: setting a feature similarity threshold, and screening suspected features whose feature similarity reaches the feature similarity threshold as the first feature.
[0050] Step S315: Find the feature level corresponding to the first feature, and count the number of first features at different feature levels And the corresponding feature similarity is , where i is the number of the levels corresponding to different main features, and j is the feature similarity number corresponding to the first feature.
[0051] Step S316, comparing the suspected features with the auxiliary features one by one, and counting the number AL of suspected features whose feature similarity reaches the feature similarity threshold.
[0052] Step S317, according to the similarity correlation function Calculate the video similarity AP, where n is The maximum number of m, the maximum number of levels corresponding to different main features, is , is the scaling factor and is greater than 0.
[0053] In actual application, the main features play a greater role, and the auxiliary features can only play an auxiliary role. The higher the feature similarity, the more it can be confirmed that the suspected target has the feature. The more similar features there are, the more similar the suspected target is to the tracking target, and the greater the possibility that it is the tracking target. Similarly, the higher the level of the suspected feature, the more it can be said that the suspected target is the tracking target. For example, the feature level is divided into level one, level two...level ten, and the feature weight ratio increases by 5%, 10%...50% respectively. Now suspected target A has two level one features with similarities of 60% and 80% respectively, one level five feature with a similarity of 90%, and the number of similarities with auxiliary features is 2. , Taking the values as 0.9 and 0.1 respectively, the video similarity is AP=0.9×[(60%×5%+80%×5%)×2+90%×25%×1]+0.1×2=0.5285. The higher the similarity, the greater the role of the surveillance video.
[0054] The steps of scanning the primary surveillance video and comparing it according to the target features to obtain the suspected target in the primary surveillance video are as follows: Step S3121, scan the primary surveillance video and extract video information in the primary surveillance video, where the video information includes video brightness, video weather, and video background color.
[0055] Step S3122, obtaining standard information of target features, the standard information including standard brightness, standard weather and standard background color.
[0056] Step S3123, comparing the video information and the standard information to see if they are consistent. If the video information and the standard information are inconsistent, adjusting the target features according to the changes in the video information to obtain the comparison features.
[0057] Step S3124: If the video information is consistent with the standard information, the target feature is directly used as a comparison feature.
[0058] Step S3125, searching the primary surveillance video for targets with similar features to the target as suspected targets.
[0059] In actual application, when making a comparison, it is necessary to consider the background situation of the surveillance video. If all surveillance videos are compared with the target features, there will be misdetection and omission. The target features will change slightly due to light brightness, weather, and background color. At this time, it is easy to get wrong results when using a unified target feature for comparison. For example, suppose we have an image of a vehicle taken in the evening. Due to the dim light, the outline and details of the vehicle are not clear enough. When the target features are obtained during the day when there is sufficient light, it is easy to get a low similarity when making a comparison. In order to make a more accurate comparison, we can use image processing software to adjust the brightness of the image, lower the brightness of the target features, and obtain the characteristics of the target features in this environment for subsequent comparison and analysis. In the weather situation, it is necessary to consider the interference caused by the weather. For example, when tracking a golden retriever, it is snowing. The target feature is originally yellow hair all over the body, but because of the snow, there will be white snowflakes on the golden retriever. Therefore, it is easy to miss the inspection when comparing it based on the yellow hair all over the body. According to whether the target feature changes under this factor, the comparison feature is adjusted to improve the accuracy of the screening process.
[0060] The steps of extracting the tracking target behavior of the intermediate surveillance video and screening the surveillance videos that meet the logic and interaction conditions according to the tracking target behavior to form an advanced surveillance video library are as follows: Step S41, extracting the moving trajectory of the tracking target in the intermediate monitoring video, and determining whether the moving trajectory follows the laws of physics.
[0061] Step S42: If the moving trajectory does not follow the laws of physics, it is determined that the corresponding intermediate monitoring video does not meet the requirements.
[0062] Step S43, obtaining the interaction between the tracking target and other elements in the intermediate monitoring video, and judging whether the interaction conforms to scientific common sense. If the interaction conforms to scientific common sense, judging that the corresponding intermediate monitoring video does not meet the requirements.
[0063] Step S44, filtering and removing the intermediate surveillance videos that do not meet the requirements from the intermediate surveillance video library to obtain the advanced surveillance video library.
[0064] In actual use, sometimes different species are too similar to each other, which may lead to misdetection. In this case, we need to judge whether it is the tracking target based on the physical laws. For example, we are tracking a golden retriever and find a suspected target, but the moving trajectory of the suspected target is from the ground to the air. Obviously, the puppy cannot fly. Then it may be a similar balloon flying into the sky, so we can filter out the video. The tracking target will interact with the surrounding objects during the tracking process, and invalid videos can be filtered out based on the interactive relationship. For example, in the surveillance video, the golden retriever was splashed with a basin of water, but it sat there motionless. According to scientific common sense, we know that the instinctive reaction of the dog when splashed with water is to avoid or run away. Therefore, we believe that the video may be a similar puppy sculpture, which does not conform to scientific common sense, so we can filter out the video. Through physical rules and interactive situations, we can exclude some invalid videos and improve the efficiency of surveillance video processing.
[0065] Based on the advanced monitoring video library, the target position of the tracking target is obtained, a three-dimensional spatial map of the target feature is formed according to the target position, and the advanced monitoring video screen is adjusted according to the three-dimensional spatial map to obtain the steps of the 3D monitoring video library, specifically: Step S51, obtaining the monitoring range of all advanced monitoring videos, and extracting the target position of the tracking target within the monitoring range.
[0066] Step S52: Classify the advanced surveillance videos with the same target location into one category and record them as the same-location surveillance videos.
[0067] Step S53: extracting the monitoring picture of the tracking target at the same time in the monitoring video at the same position to form a three-dimensional spatial map of the tracking target.
[0068] Step S54, obtaining the relative orientation of the tracking target and other objects in the monitoring video according to the three-dimensional space diagram, and reflecting the spatial position of the tracking target by adjusting the relative sizes of the tracking target and other objects.
[0069] Step S55, counting the order of distances between the tracking target and other objects and the monitoring device, increasing the color brightness of the tracking target and other objects in order from far to near according to the order of distances, and obtaining a 3D monitoring video library.
[0070] In actual use, the better the 3D effect, the clearer the position of the tracking target and other objects, thus obtaining a more accurate position. By shooting pictures of the tracking target at the same position and different positions, a three-dimensional spatial map of the tracking target can be formed. According to the three-dimensional spatial map of the tracking target, the position relationship between the tracking target and other objects can be obtained in a more three-dimensional manner. Larger objects are usually perceived as closer visually, while smaller objects are farther away. By appropriately adjusting the relative size of the tracking target and other objects, users can better know the position of the tracking target. In real life, objects near are usually brighter and more colorful than objects far away. This effect can be simulated by changing the color and brightness, thereby increasing the sense of depth and improving the 3D effect. For example, when tracking a puppy, because of its small size, it is beneficial to quickly capture it on the spot after obtaining the exact position in the monitoring. In addition, some surveillance videos will cause visual dislocation due to the shooting angle. By enhancing the 3D effect, the visual dislocation can be effectively reduced.
[0071] The steps of obtaining the importance of information in different areas of the surveillance video, filtering according to the importance of the information, obtaining valid videos and pushing them to the user end are as follows: Step S61, performing area division according to objects in the surveillance video to obtain different object areas.
[0072] Different objects are divided into one area. For example, if there are grass, people, dogs and balloons in the surveillance video, four areas are divided, and the main objects in the area are grass, people, dogs and balloons respectively.
[0073] Step S62: extract the object area features, search for similar surveillance videos in the surveillance data, and count the number DS of similar surveillance videos.
[0074] Step S63: based on the interaction situation, extract the interaction times DC and interaction times DJ between different objects and the tracking target.
[0075] Step S64: according to the information importance related function Calculate the information importance DZ, where , , is the scaling factor and is greater than 0.
[0076] Step S65: The information importance of the suspected target area is the highest value, and the corresponding filtering degree is set according to the information importance. After filtering different object areas, valid videos are obtained and pushed to the user end.
[0077] In practical applications, image filtering refers to processing the pixel values of an image through specific algorithms and techniques to achieve the purpose of improving image quality, enhancing specific information or extracting image features. However, image filtering consumes a lot of resources, so unimportant areas can be left unfiltered or lightly filtered to improve image processing efficiency and reduce resource consumption. When the object features are more difficult to distinguish from other objects, more similar surveillance videos will be found, indicating that the amount of information carried is less, the information is less important, and no excessive processing is required. However, if the object interacts with the tracking target more times and for a longer time, it can provide more effective information. For example, the tracking target is a puppy. There are puppies, grass, shops and platforms in the surveillance video. Grass can be seen everywhere and there is no need to process it because it does not provide effective information. It is also easy to find many shops with repeated store names, so only slight processing is required. The platform can reveal the specific location, so the filtering degree is greater. As the main tracking object, the puppy needs to be carefully filtered so that the user can confirm whether it is the tracking target. However, if the interaction between the grass and the puppy is long, the filtering level needs to be increased. Since the interaction between the grass and the puppy may leave hair and other items, which is conducive to detection, it is necessary to determine the specific grass conditions. While filtering, unnecessary processing is reduced to improve the processing speed of surveillance video.
[0078] A video surveillance processing system, by applying a video surveillance processing method as described above, a target feature module obtains at least one feature of a tracking target and records it as a target feature, extracts the most recent appearance information of the tracking target and records it as the last information.
[0079] The primary screening module obtains the monitoring data of the video surveillance, and screens out the monitoring videos that meet the regional requirements from the monitoring data according to the latest information to form a primary monitoring video library.
[0080] The intermediate screening module sets an information volume threshold based on target features and selects surveillance videos that reach the information volume threshold from the primary surveillance video library to form an intermediate surveillance video library.
[0081] The advanced screening module extracts the tracking target behavior of the intermediate surveillance video, and screens the surveillance videos that meet the logic and interaction conditions according to the tracking target behavior to form an advanced surveillance video library.
[0082] The 3D adjustment module obtains the target position of the tracking target based on the advanced monitoring video library, forms a three-dimensional space map of the target characteristics according to the target position, adjusts the advanced monitoring video picture according to the three-dimensional space map, and obtains the 3D monitoring video library.
[0083] The filtering and pushing module obtains the importance of information in different areas of the surveillance video, filters it according to the importance of the information, obtains valid video and pushes it to the user end.
[0084] In actual application, a video surveillance processing system is formed by setting the functions of each module. The video surveillance processing system is used to screen, filter and enhance the 3D effect of the surveillance video, which is conducive to obtaining clearer and more accurate surveillance video and improving tracking efficiency.
[0085] The implementation principle of this system is: first, use the target feature module to obtain at least one feature of the tracking target and record it as the target feature, extract the latest appearance information of the tracking target and record it as the last information. And according to whether the target feature can be changed, it is divided into main features and auxiliary features, and different feature levels are divided according to the classification units of the category to which the main feature belongs. The primary screening module obtains the monitoring data of the video surveillance, and selects the monitoring videos that meet the area that the tracking target can reach from the monitoring data according to the last information and the traffic route map to form a primary monitoring video library. The intermediate screening module determines the information volume of the monitoring video based on the monitoring similarity of the monitoring video after the tracking target is compared, combined with the duration and area of the tracking target in the monitoring video. Set the information volume threshold, select the monitoring video that reaches the information volume threshold from the primary monitoring video library to form an intermediate monitoring video library. The advanced screening module extracts the tracking target behavior of the intermediate monitoring video, and selects the monitoring video that conforms to the logic and interaction according to the tracking target behavior to form an advanced monitoring video library. The 3D adjustment module obtains the target position of the tracking target based on the advanced surveillance video library, forms a three-dimensional spatial map of the target features according to the target position, adjusts the advanced surveillance video screen according to the three-dimensional spatial map, and obtains the 3D surveillance video library. The filtering and pushing module obtains the importance of information in different areas of the surveillance video, filters according to the importance of the information, obtains valid videos and pushes them to the user end.
[0086] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.
Claims
1. A video surveillance processing method, It is characterized in that The following steps are involved: Obtain at least one feature of the tracking target and record it as the target feature, extract the most recent appearance information of the tracking target and record it as the last information; Obtain monitoring data from video surveillance, and select monitoring videos that meet regional requirements from the monitoring data based on the latest information to form a primary monitoring video library; Based on the target feature, an information volume threshold is set, and surveillance videos reaching the information volume threshold are selected from the primary surveillance video library to form an intermediate surveillance video library; Extract the tracking target behavior of the intermediate surveillance video, and filter the surveillance videos that meet the logic and interaction conditions according to the tracking target behavior to form an advanced surveillance video library; Based on the advanced surveillance video library, a target position of the tracking target is obtained, a three-dimensional spatial map of the target feature is formed according to the target position, and the advanced surveillance video picture is adjusted according to the three-dimensional spatial map to obtain a 3D surveillance video library; Obtain the importance of information in different areas of the surveillance video, filter it according to the importance of the information, obtain valid video and push it to the user end.
2. A video surveillance processing method according to claim 1, It is characterized in that The step of obtaining at least one feature of the tracking target and recording it as the target feature, and extracting the most recent appearance information of the tracking target and recording it as the last occurrence information, specifically comprises: Obtain at least one feature of the tracking target and record it as the target feature, determine whether the target feature is a fixed feature, and if the target feature is a fixed feature, classify the target feature as a main feature; If the target feature is not a fixed feature, the target feature is classified as an auxiliary feature; Get the category to which the tracking target belongs, and form a corresponding level according to the classification unit of the category; Match the classification unit to which the main feature belongs, find the level corresponding to the classification unit, and obtain the feature level of the main feature; The last information includes the last location and the last time point.
3. A video surveillance processing method according to claim 2, It is characterized in that The step of obtaining the monitoring data of the video monitoring and selecting the monitoring videos meeting the regional requirements from the monitoring data according to the last information to form a primary monitoring video library is specifically as follows: Get the real-time point, and calculate the time difference between the real-time point and the appearance time point; Obtain a traffic operation route map, and determine the maximum moving distance of the tracking target based on the traffic operation route map and the time difference; With the last location as the center and the maximum moving distance as the radius, a search area is formed; Extract the monitoring locations of all monitoring videos from the monitoring data and determine whether the monitoring locations are within the search area; If the monitoring location is within the search area, the corresponding monitoring video will form a primary monitoring video library.
4. A video surveillance processing method according to claim 3, It is characterized in that The steps of setting an information volume threshold based on the target feature and selecting surveillance videos reaching the information volume threshold from the primary surveillance video library to form an intermediate surveillance video library are specifically as follows: Compare the target features with the primary surveillance video to obtain the video similarity AP; Extract the duration BS of the target being tracked in the primary surveillance video, and extract the ratio BB of the area of the target being tracked in the primary surveillance video to the area of the target being tracked; According to the information related function The information volume CX is calculated, where , , is the scale factor and is greater than 0; An information volume threshold is set, and surveillance videos that reach the information volume threshold are selected from the primary surveillance video library to form an intermediate surveillance video library.
5. A video surveillance processing method according to claim 4, It is characterized in that The steps of comparing the target features with the primary surveillance video to obtain the video similarity AP are as follows: Set the feature weight ratio of different main features according to the feature level , where i is the number of the corresponding level of different main characteristics; Scan the primary surveillance video, compare it according to the target features, and obtain the suspected target in the primary surveillance video; Extract the suspected features of the suspected target, compare the suspected features with the main features one by one, and obtain the feature similarity between the suspected features and the main features; Set a feature similarity threshold, and select suspected features whose feature similarity reaches the feature similarity threshold as the first feature; Find the feature level corresponding to the first feature and count the number of first features at different feature levels And the corresponding feature similarity is , where i is the number of the levels corresponding to different main features, and j is the feature similarity number corresponding to the first feature; Compare the suspected features with the auxiliary features one by one, and count the number AL of suspected features whose feature similarity reaches the feature similarity threshold; According to the similarity correlation function Calculate the video similarity AP, where n is The maximum number of m, the maximum number of levels corresponding to different main features, is , is the scaling factor and is greater than 0.
6. A video surveillance processing method according to claim 5, It is characterized in that The step of scanning the primary surveillance video and comparing it according to the target features to obtain the suspected target in the primary surveillance video is specifically as follows: Scanning the primary surveillance video to extract video information in the primary surveillance video, wherein the video information includes video brightness, video weather, and video background color; Acquire standard information of target features, wherein the standard information includes standard brightness, standard weather, and standard background color; Compare the video information and the standard information to see if they are consistent. If they are inconsistent, adjust the target features according to the changes in the video information to obtain the comparison features. If the video information is consistent with the standard information, the target feature is directly used as the comparison feature; Search for targets with similar features to the target in the primary surveillance video as suspected targets.
7. A video surveillance processing method according to claim 6, It is characterized in that The steps of extracting the tracking target behavior of the intermediate monitoring video and screening the monitoring videos that meet the logic and interaction conditions according to the tracking target behavior to form the advanced monitoring video library are specifically as follows: Extract the moving trajectory of the target in the intermediate surveillance video and determine whether the moving trajectory follows the laws of physics; If the movement trajectory does not follow the laws of physics, the corresponding intermediate surveillance video is judged to be unqualified; Obtain the interaction between the tracking target and other elements in the intermediate surveillance video, and determine whether the interaction conforms to scientific common sense. If the interaction does not conform to scientific common sense, the corresponding intermediate surveillance video is determined to be unqualified. Intermediate surveillance video libraries that do not meet the requirements are screened and removed to obtain advanced surveillance video libraries.
8. A video surveillance processing method according to claim 7, It is characterized in that Based on the advanced surveillance video library, the target position of the tracking target is obtained, a three-dimensional spatial map of the target feature is formed according to the target position, and the advanced surveillance video picture is adjusted according to the three-dimensional spatial map to obtain the 3D surveillance video library, specifically, the steps are: Obtain the monitoring range of all advanced monitoring videos and extract the target position of the tracking target within the monitoring range; Advanced surveillance videos with the same target location are grouped into one category and recorded as same-location surveillance videos; Extract the monitoring picture of the tracking target at the same time in the monitoring video of the same location to form a three-dimensional spatial map of the tracking target; The relative position of the tracking target and other objects in the surveillance video is obtained according to the three-dimensional space diagram, and the spatial position of the tracking target is reflected by adjusting the relative size of the tracking target and other objects; The order of the distances of the tracking target and other objects from the monitoring device is counted, and the color brightness of the tracking target and other objects is increased from far to near according to the order of distances, and a 3D monitoring video library is obtained.
9. A video surveillance processing method according to claim 8, It is characterized in that The steps of obtaining the importance of information in different areas of the surveillance video, filtering according to the importance of the information, obtaining valid videos and pushing them to the user end are as follows: Divide the area according to the objects in the surveillance video to obtain different object areas; Extract the object area features to find similar surveillance videos in the surveillance data and count the number of similar surveillance videos DS; Based on the interaction situation, extract the interaction times DC and interaction times DJ between different objects and the tracking target; According to the importance of information Calculate the information importance DZ, where , , is the scale factor and is greater than 0; The information importance of the suspected target area is the highest value. The corresponding filtering degree is set according to the information importance. After filtering different object areas, valid videos are obtained and pushed to the user end.
10. A video surveillance processing system, It is characterized in that By applying a video surveillance processing method as described in any one of claims 1 to 9, comprising: The target feature module obtains at least one feature of the tracking target and records it as the target feature, extracts the most recent appearance information of the tracking target and records it as the last occurrence information; The primary screening module obtains the monitoring data of the video surveillance, and screens the monitoring videos that meet the regional requirements from the monitoring data according to the last information to form a primary monitoring video library; The intermediate screening module sets an information volume threshold based on the target feature, and selects surveillance videos that reach the information volume threshold from the primary surveillance video library to form an intermediate surveillance video library; The advanced screening module extracts the tracking target behavior of the intermediate surveillance video, and screens the surveillance videos that meet the logic and interaction conditions according to the tracking target behavior to form an advanced surveillance video library; A 3D adjustment module, based on the advanced surveillance video library, obtains a target position of a tracking target, forms a three-dimensional spatial map of target features according to the target position, and adjusts the advanced surveillance video picture according to the three-dimensional spatial map to obtain a 3D surveillance video library; The filtering and pushing module obtains the importance of information in different areas of the surveillance video, filters it according to the importance of the information, obtains valid video and pushes it to the user end.
Citation Information
Patent Citations
Trajectory generation method and device for query object
CN107315755A
Visitor prewarning method and system based on face recognition
CN108537922A
Pedestrian recognition system, recognition method and computer readable storage medium
CN109934176A
Data analysis method and device, electronic equipment and computer storage medium
CN110705476A
Video-based target association method and device and readable storage medium
CN114219828A
