Alarm method and alarm system for intelligent video monitoring

Abstract

The invention discloses an alarm method and an alarm system for intelligent video monitoring, and relates to the technical field of computer vision. The method comprises the following steps: acquiring a real-time video stream and a target detection model corresponding to a monitoring area, and detecting the real-time video stream based on the target detection model to obtain a moving target included in the real-time video stream; extracting motion data of the moving target, tracking the moving target in real time according to the motion data of the moving target, and mapping the moving target to a 3D panoramic monitoring picture corresponding to the monitoring area and displaying the moving target; and acquiring a preset alarm rule, tracking and detecting the real-time action of the moving target according to the preset alarm rule, and if the real-time action of the moving target is detected to meet the preset alarm rule, triggering a linkage alarm response corresponding to the preset alarm rule. The panoramic view is generated by fusing the multi-camera data, and the early warning function of the system is realized by combining the multi-dimensional analysis of the target type, the speed and the track, so that the monitoring effectiveness is greatly improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer vision technology, and in particular to an alarm method and an alarm system for intelligent video surveillance. Background Art

[0002] Traditional video surveillance systems face multiple technical bottlenecks during large-scale deployment and functional upgrades: First, with the surge in monitoring points and the expansion of coverage, system complexity and data volume have grown exponentially, resulting in a significant increase in security labor costs and management pressure; second, existing technologies rely on passive security modes such as manual patrols and post-event video backtracking, which makes it difficult to achieve real-time identification of abnormal behavior, and multi-camera data is isolated and scattered, lacking panoramic situational awareness capabilities; third, traditional infrared sensors and other sensors are easily affected by climate and environment, resulting in a high false alarm rate, and the existing target detection system does not integrate 3D spatial mapping technology, resulting in the inability to accurately project the target position and motion trajectory into the virtual environment. At the same time, the single-dimensional alarm rule design is difficult to coordinate multi-modal characteristics such as target type, speed, and trajectory, resulting in limited warning accuracy and response efficiency.

[0003] Therefore, how to effectively collect video data information and conduct effective intelligent prevention and control is becoming more and more important and urgent. Summary of the invention

[0004] The purpose of this application is to solve at least one of the above-mentioned technical deficiencies.

[0005] On the one hand, an embodiment of the present application provides an alarm method for intelligent video surveillance, the method comprising:

[0006] Acquire the real-time video stream and target detection model corresponding to the monitoring area, and detect the real-time video stream based on the target detection model to obtain the moving target included in the real-time video stream;

[0007] Extract the motion data of the moving target, track the moving target in real time according to the motion data of the moving target, and map the moving target to the 3D panoramic monitoring screen corresponding to the monitoring area and display it;

[0008] The preset alarm rules are obtained, and the real-time motion tracking and detection of the mobile target is carried out according to the preset alarm rules. If the real-time motion of the mobile target is detected to meet the preset alarm rules, the linkage alarm response corresponding to the preset alarm rules is triggered.

[0009] Optionally, the real-time video stream is detected based on the target detection model to obtain the moving target included in the real-time video stream, including:

[0010] Based on the adaptive Gaussian mixture model, the background of the real-time video stream is dynamically modeled to obtain the modeled real-time video stream;

[0011] Perform optical flow processing on the modeled real-time video stream to obtain the inter-frame motion vector field corresponding to the real-time video stream;

[0012] Detect the region of the inter-frame motion vector field based on the object detection model to obtain the moving objects included in the real-time video stream.

[0013] Optionally, detecting the region of the inter-frame motion vector field based on the object detection model to obtain the moving objects included in the real-time video stream includes:

[0014] Detect the region of the inter-frame motion vector field based on the object detection model to obtain at least one initial moving object;

[0015] Adopt a multi-level joint filtering mechanism to filter at least one initial moving object to obtain the moving objects included in the real-time video stream. The multi-level joint filtering mechanism includes at least one of tracking type, object type, size threshold, and confidence threshold.

[0016] Optionally, the adaptive Gaussian mixture model performs dynamic modeling based on the following formula:

[0017]

[0018] where X t is the current pixel value, K is the number of Gaussian distributions, is the weight, u k,t and, ∑ k,t are the mean and covariance matrix α base is the basic learning rate, ΔI t is the average pixel difference between the current frame and the background model, ΔS t is the average optical flow amplitude between adjacent frames, I max is the normalization coefficient of the pixel difference, S max is the normalization coefficient of the optical flow amplitude, M k,t is the binary indicator function.

[0019] Optionally, the motion data includes position and velocity. Real-time tracking of the moving object according to the motion data of the moving object includes:

[0020] According to the position and velocity of the moving object, determine at least one predicted velocity and at least one predicted position of the moving object based on the Kalman filter algorithm;

[0021] Perform trajectory smoothing processing on at least one predicted position of the moving object to obtain at least one processed predicted position, and obtain the predicted motion trajectory of the moving object according to at least one processed predicted position and at least one predicted velocity;

[0022] Determine the identifier of the moving target, and associate the predicted movement trajectory of the moving target with the identifier to achieve real-time tracking of the moving target.

[0023] Optionally, the predicted position is represented based on pixel coordinates, and the moving target is mapped to and displayed in the 3D panoramic surveillance image corresponding to the monitoring area, including:

[0024] Obtain the installation parameters of the camera corresponding to the monitoring area and the 3D virtual target model corresponding to the moving target. The installation parameters include installation height, pitch angle, and field of view;

[0025] Perform coordinate transformation on the pixel coordinates of the predicted position based on the installation parameters of the camera to obtain the three-dimensional geographical coordinates corresponding to the predicted position;

[0026] Based on the three-dimensional geographical coordinates corresponding to the predicted position, map the 3D virtual target model to the pixel coordinates of the 3D panoramic surveillance image and display it. The 3D panoramic surveillance image is established after calibrating the installation height, pitch angle, and field of view of the camera according to the virtual character model and virtual scale.

[0027] Optionally, the preset alarm rules include at least one of a target leaving the post alarm rule, a sleeping on the post alarm rule, an entering alarm rule, a leaving alarm rule, a target appearance alarm rule, a target disappearance alarm rule, a target classification alarm rule, a speed alarm rule, a direction alarm rule, an occlusion alarm rule, a tailing alarm rule, a flame alarm rule, a color alarm rule, an abandoned object alarm rule, an object moving alarm rule, and a traffic statistics alarm rule.

[0028] Optionally, triggering the linkage alarm response corresponding to the preset alarm rule includes:

[0029] Determine at least one linkage alarm device corresponding to the preset alarm rule, and send an alarm instruction to at least one linkage alarm device;

[0030] Record the accurate position and corresponding real-time video stream of the moving target when the linkage alarm response is triggered, and generate a record log by associating the accurate position and corresponding real-time video stream of the moving target with the moving target management storage.

[0031] On the other hand, an embodiment of the present application provides an alarm system for intelligent video surveillance. The alarm system includes:

[0032] A video acquisition module, configured to support accessing a camera according to a set protocol and obtain a real-time video stream corresponding to the monitoring area through hardware decoding;

[0033] Target detection and tracking module: It is used to detect the real-time video stream based on the target detection model, obtain the moving targets included in the real-time video stream, extract the motion data of the moving targets, and track the moving targets in real time according to the motion data of the moving targets;

[0034] 3D mapping display module: It is used to obtain the installation parameters of the cameras corresponding to the monitoring area and the 3D virtual target model corresponding to the moving targets, perform coordinate conversion on the pixel coordinates of the predicted position based on the installation parameters of the cameras to obtain the three-dimensional geographical coordinates corresponding to the predicted position, and map the 3D virtual target model into the pixel coordinates of the 3D panoramic monitoring screen and display it based on the three-dimensional geographical coordinates corresponding to the predicted position;

[0035] Behavior analysis and alarm module: It is used to configure expandable preset alarm rules, perform real-time action tracking detection on the moving targets according to the preset alarm rules, and trigger a linkage alarm response and generate a record log when it is detected that the real-time actions of the moving targets meet the preset alarm rules.

[0036] Optionally, the alarm system is also configured with an SDK docking module. The SDK docking module includes an SDK development kit, preset interfaces, and example programs of the linkage subsystem. The SDK development kit includes a standard network encapsulation protocol, event stream operations, preset interfaces, and source code of the development document example programs.

[0037] On the other hand, an embodiment of the present application provides an electronic device, including a processor and a memory:

[0038] The memory is configured to store machine-readable instructions. When the instructions are executed by the processor, the processor executes any one of the methods in an intelligent video surveillance alarm method.

[0039] The beneficial effects brought by the technical solutions provided by the embodiments of the present application at least include:

[0040] In the embodiments of the present application, for the monitoring area, the corresponding video stream can be obtained and the 3D panoramic monitoring corresponding to the monitoring area can be generated, and then the real-time actions of the moving targets in the video stream are tracked and detected in combination with the preset alarm rules, and when the preset alarm rules are met, the linkage alarm response corresponding to the preset alarm rules is triggered. That is to say, in the present application, a panoramic view is generated by fusing multi-camera data, and early warning analysis is performed by combining multi-dimensional data such as the target type, speed, and trajectory of the moving targets with the alarm rules, realizing the early warning function of the system, changing the traditional video surveillance method of relying on human eyes to stare at the electronic eyes and rolling inspections of the multi-camera pictures, and at the same time greatly improving the monitoring effectiveness, and being more applicable to complex security monitoring scenarios.

[0041] In the embodiments of the present application, when dynamically modeling the background of a real-time video stream, the learning rate can be dynamically adjusted according to the amount of light change and scene dynamics, thus solving the problem of model lag of traditional MOG2 under sudden light changes and dynamic scenes. Moreover, through the merging and splitting of Gaussian distributions, the modeling accuracy and computational load are balanced, thereby realizing the adaptive update of the background model in complex environments and significantly improving the robustness of moving target detection.

[0042] In the embodiments of the present application, computational load can be reduced through optical flow processing. At the same time, through a multi-level joint filtering mechanism, the number of false alarms can be reduced under complex scenes (such as shaking leaves and light changes), and irrelevant detection frames can be filtered out, saving subsequent trajectory tracking and optimizing resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0044] Figure 1 It is a schematic flowchart of an alarm method for intelligent video surveillance provided by the embodiments of the present application;

[0045] Figure 2 It is a schematic diagram of the category of a moving target provided by the embodiments of the present application;

[0046] Figure 3 It is a schematic diagram of the setting of a target type provided by the embodiments of the present application;

[0047] Figure 4 It is a schematic diagram of the setting of a filtering method and a tracking target type provided by the embodiments of the present application;

[0048] Figure 5 It is a schematic diagram of mapping a moving target to a 3D panoramic surveillance screen provided by the embodiments of the present application;

[0049] Figure 6 It is a schematic diagram of a 3D virtual scene provided by the embodiments of the present application;

[0050] Figure 7 It is a schematic diagram of the setting of an alarm rule provided by the embodiments of the present application;

[0051] Figure 8 It is a schematic diagram of a topological network provided by the embodiments of the present application;

[0052] Figure 9A schematic diagram of a perimeter visualization intelligent early warning scenario provided by an embodiment of the present application;

[0053] Figure 10 A schematic diagram of a safe campus visualization intelligent early warning scenario provided by an embodiment of the present application;

[0054] Figure 11 A schematic diagram of a production safety visualization early warning scenario provided by an embodiment of the present application;

[0055] Figure 12 A schematic diagram of a visualization intelligent early warning scenario for the civil explosive industry provided by an embodiment of the present application;

[0056] Figure 13 A schematic diagram of a visualization intelligent early warning scenario for off - duty sleeping on the job provided by an embodiment of the present application;

[0057] Figure 14 A schematic diagram of intelligent analysis of commercial customer behavior big data provided by an embodiment of the present application;

[0058] Figure 15 A schematic diagram of the structure of an alarm system for intelligent video surveillance provided by an embodiment of the present application;

[0059] Figure 16 A schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0060] The embodiments of the present application will be described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements with the same or similar functions from beginning to end. The embodiments described below with reference to the drawings are exemplary and are only used to explain the present application and cannot be construed as a limitation of the present invention.

[0061] Those skilled in the art of the present technology can understand that unless specifically stated otherwise, the singular forms "a", "an", "the" and "said" used herein may also include the plural forms. It should be further understood that the term "including" used in the specification of the present application means the presence of the described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or their groups. It should be understood that when we say that an element is "connected" or "coupled" to another element, it can be directly connected or coupled to other elements, or there may also be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The phrase "and / or" used herein includes all or any unit and all combinations of one or more of the associated listed items.

[0062] To make the objectives, technical solutions, and advantages of this application clearer, the following will further describe the embodiments of this application in detail in conjunction with the accompanying drawings.

[0063] The following will use specific embodiments to detail the technical solutions of this application and how the technical solutions of this application solve the above technical problems. These several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this application will be described below in conjunction with the accompanying drawings.

[0064] Specifically, as Figure 1 shown, the method may include:

[0065] Step S101, obtain the real-time video stream corresponding to the monitoring area and the target detection model, and detect the real-time video stream based on the target detection model to obtain the moving targets included in the real-time video stream.

[0066] Optionally, the monitoring area refers to the area that needs to be video-monitored. There will be camera equipment (such as cameras) installed in this monitoring area, and the real-time video stream corresponding to this monitoring area can be obtained based on the installed camera equipment. Correspondingly, after obtaining the real-time video stream, the real-time video stream can be detected based on the pre-trained target detection model, and the moving targets included in the real-time video stream can be extracted. Among them, the specific types of moving targets that need to be extracted can be set according to actual needs, which can be people or objects, etc. The embodiments of this application do not limit this, for example, any one of the categories shown in 2 can be used as the moving target to be recognized. And the specific network structure of the target detection model can also be selected according to actual needs. Any network model that can implement the function of detecting the real-time video stream in this application to obtain moving targets can be used as the target detection model in this application, and the embodiments of this application do not limit this.

[0067] Step S102, extract the motion data of the moving target, and track the moving target in real time according to the motion data of the moving target, and map the moving target to the 3D panoramic monitoring screen corresponding to the monitoring area and display it.

[0068] Optionally, after determining the moving targets in the real-time video stream, the motion data of the moving target can be obtained, so that the moving target can be tracked in real time. At the same time, in order to better monitor the moving target, the moving target can be mapped to the 3D panoramic monitoring screen corresponding to the monitoring area and displayed.

[0069] In an optional embodiment of this application, detecting the real-time video stream based on the target detection model to obtain the moving targets included in the real-time video stream includes:

[0070] Dynamically model the background of the real-time video stream based on an adaptive Gaussian mixture model to obtain the modeled real-time video stream;

[0071] Perform optical flow processing on the modeled real-time video stream to obtain the inter-frame motion vector field corresponding to the real-time video stream;

[0072] Detect the region of the inter-frame motion vector field based on the target detection model to obtain the moving targets included in the real-time video stream.

[0073] Optionally, when determining the moving targets included in the real-time video stream, the background of the real-time video stream can be dynamically modeled based on the adaptive Gaussian mixture model to obtain the modeled real-time video stream, and then optical flow processing is performed on the modeled real-time video stream to obtain the inter-frame motion vector field corresponding to the real-time video stream. Among them, performing optical flow processing to obtain the inter-frame motion vector field corresponding to the real-time video stream can refer to obtaining the inter-frame motion vector field based on the method of calculating sparse optical flow (Lucas-Kanade algorithm) or dense optical flow (Farneback algorithm).

[0074] Furthermore, after obtaining the inter-frame motion vector field corresponding to the real-time video stream, a pre-trained target detection model can be obtained, and then the region of the obtained inter-frame motion vector field is detected based on the target detection model to obtain the moving targets included in the real-time video stream.

[0075] In an alternative embodiment of the present application, the adaptive Gaussian mixture model performs dynamic modeling based on the following formula:

[0076]

[0077] where X t is the current pixel value, K is the number of Gaussian distributions, is the weight, u k,t and, ∑ k,t are the mean and covariance matrix α base is the basic learning rate, ΔI t is the average pixel difference between the current frame and the background model, ΔS t is the average value of the optical flow amplitude between adjacent frames, I max is the normalization coefficient of the pixel difference, S max is the normalization coefficient of the optical flow amplitude, M k,t is the binary indicator function.

[0078] Optionally, in the present application, ΔI t can measure the overall difference between the current frame and the background model. For example, when there is a sudden change in illumination (such as a light being turned on or clouds blocking sunlight), ΔI tIt will increase significantly, and the learning rate will also increase accordingly, accelerating the update of the background model and avoiding false foreground detection caused by light changes. ΔS t Calculate the average motion amplitude between adjacent frames through the optical flow method to reflect the dynamic degree of the scene (such as leaf shaking and crowded areas). When the scene has high dynamicity, ΔS t increases, and the learning rate decreases, thereby preventing the background model from overfitting dynamic interference due to frequent motion. The normalization coefficient I max and S max Scale ΔI t and ΔS t to the interval [0, 1] to prevent the learning rate from getting out of control due to different parameter dimensions. Among them, the lower limit of the learning rate is The upper limit is In extreme cases, the learning rate is restricted to a reasonable range (such as 0.01 ≤ learning rate ≤ 0.2), which can avoid the model from updating too fast or stagnating.

[0079] In the embodiment of the present application, when dynamically modeling the background of a real-time video stream, the learning rate can be dynamically adjusted according to the amount of light change and the scene dynamicity, thereby solving the problem of model lag of traditional MOG2 in the case of sudden light changes and dynamic scenes, and balancing the modeling accuracy and computational load through Gaussian distribution merging and splitting, so as to achieve adaptive update of the background model in a complex environment and significantly improve the robustness of moving target detection.

[0080] In an optional embodiment of the present application, based on the target detection model, the region of the inter-frame motion vector field is detected to obtain the moving targets included in the real-time video stream, including:

[0081] Based on the target detection model, the region of the inter-frame motion vector field is detected to obtain at least one initial moving target;

[0082] A multi-level joint filtering mechanism is used to filter at least one initial moving target to obtain the moving targets included in the real-time video stream. The multi-level joint filtering mechanism includes at least one of the tracking type, target type, size threshold, and confidence threshold.

[0083] Optionally, after obtaining the region of the inter-frame motion vector field, at this time, the region of the inter-frame motion vector field can be detected based on the target detection model to obtain at least one initial moving target. In order to improve the accuracy of moving target detection, at this time, a multi-level joint filtering mechanism can be used to filter the obtained at least one initial moving target, and the filtered moving target is used as the moving target included in the real-time video stream in the present application.

[0084] Among them, the multi-level joint filtering mechanism includes at least one of a tracking type, a target type, a size threshold, and a confidence threshold. The tracking type refers to the target type to be tracked set through a configuration window based on the analysis algorithm rules. When "include" is selected, the tracking type is that the set target type is tracked; when "exclude" is selected, the tracking type is that the set target type is not tracked. The target type refers to the type of moving target to be detected, and the identifier (i.e., name) of each target type can be determined according to the target size and target speed. For example, it can be set through the Figure 3 interface. The size threshold refers to filtering the size of the detected target. Specifically, the minimum detection box size (such as 20×20 pixels) can be defined. At this time, false detections caused by noise or small objects (flies, raindrops) can be filtered. The maximum detection box size (such as covering 80% of the screen area) can be defined to exclude abnormally large boxes caused by background modeling errors. The confidence threshold refers to filtering based on the confidence threshold. Specifically, the confidence threshold can be dynamically adjusted. The default threshold is 0.5 (this value can balance the recall rate and precision). In low-light or blurred scenarios, the threshold is reduced to 0.3 to avoid missed detections. Optionally, in practice, the filtering method and the tracking target type can be set through the Figure 4 interface shown.

[0085] In the embodiment of the present application, the computational load can be reduced through stream processing. At the same time, through the multi-level joint filtering mechanism, the number of false alarms can be reduced in complex scenarios (such as leaf shaking, light change), and irrelevant detection boxes can be filtered, saving subsequent trajectory tracking and optimizing resource utilization.

[0086] In an alternative embodiment of the present application, the motion data includes position and speed. The moving target is tracked in real time according to the motion data of the moving target, including:

[0087] Based on the position and speed of the moving target, at least one predicted speed and at least one predicted position of the moving target are determined based on the Kalman filtering algorithm;

[0088] The trajectory of at least one predicted position of the moving target is smoothed to obtain at least one processed predicted position, and based on the at least one processed predicted position and at least one predicted speed, the predicted motion trajectory of the moving target is obtained;

[0089] The identifier of the moving target is determined, and the predicted motion trajectory of the moving target is associated with the identifier to achieve real-time tracking of the moving target.

[0090] Optionally, the bounding box (BoundingBox) of the moving target can be obtained through an object detection model (such as the YOLO model), and the pixel coordinates (x, y) of its center point can be extracted. Then, the internal parameters (focal lengths fx, fy, distortion coefficients) and external parameters (installation height H, pitch angle θ) of the camera can be obtained using the checkerboard calibration method. The pixel coordinates are converted into world coordinates (X, Y) through the pre-computed homography matrix H. Then, the depth information (such as binocular cameras or LiDAR) is combined to calculate the three-dimensional coordinates (X, Y, Z), and the position of the moving target is represented by the three-dimensional coordinates.

[0091] Furthermore, the world coordinates of the same target in two consecutive frames are obtained as (Xt-1, Yt-1) and (Xt, Yt) respectively. At this time, the displacement Δd = √((Xt - Xt-1)² + (Yt - Yt-1)²) can be obtained, and then the real-time speed of the moving target can be obtained based on the obtained displacement.

[0092] Optionally, an identifier can be assigned to each moving target, and the historical position sequence of the moving target is bound and stored with its label. Then, a sliding window (such as the last 30 frames) is used to limit the trajectory length to avoid memory overflow. Correspondingly, at least one predicted speed and at least one predicted position of the moving target are determined using Kalman filtering (or an improved filtering algorithm) based on the position and speed of the moving target. Further, at least one predicted position of the moving target is smoothed to obtain at least one processed predicted position, and then the predicted motion trajectory of the moving target is obtained based on at least one processed predicted position and at least one predicted speed.

[0093] Furthermore, the object detection model is run on the current frame to output the bounding boxes and categories of all candidate targets of the moving target, and the bounding boxes and categories of all candidate targets are filtered. Then, at least one predicted position predicted based on Kalman filtering (or an improved filtering algorithm) is associated with the identifier of the moving target. Among them, when making the association, the matching cost between the detection box and the prediction box can be calculated (for example, the spatial overlap degree can be measured based on IoU (Intersection over Union), the motion consistency can be ensured through the cosine similarity of the speed direction, the feature vector can be extracted through the Re-ID model, and the cosine distance is calculated to characterize the appearance feature similarity). Then, the Hungarian algorithm is used to solve the optimal matching to ensure that each detection box is uniquely associated with the prediction box. Finally, for the successfully matched tracking target, its Kalman filtering state is updated to obtain the next predicted position and associated with the identifier, so as to realize the real-time tracking of the moving target.

[0094] In an optional embodiment of the present application, the predicted position is represented based on pixel coordinates, and the moving target is mapped to the corresponding 3D panoramic monitoring screen of the monitoring area and displayed, including:

[0095] Obtain the installation parameters of the cameras corresponding to the monitoring area and the 3D virtual target model corresponding to the moving target. The installation parameters include the installation height, the pitch angle, and the field of view.

[0096] Perform coordinate transformation on the pixel coordinates of the predicted position based on the installation parameters of the camera to obtain the three-dimensional geographical coordinates corresponding to the predicted position.

[0097] Based on the three-dimensional geographical coordinates corresponding to the predicted position, map the 3D virtual target model into the pixel coordinates of the 3D panoramic monitoring screen and display it. The 3D panoramic monitoring screen is established after calibrating the installation height, the pitch angle, and the field of view of the camera according to the virtual human model and the virtual scale.

[0098] Optionally, for each type of moving target, a corresponding 3D virtual target model can be preset. Correspondingly, when performing real-time monitoring of the monitoring area, the installation parameters of the cameras corresponding to the monitoring area can be obtained. The installation parameters of the camera can specifically include the installation height, the pitch angle, and the field of view of the camera. Then, perform coordinate transformation according to the obtained installation parameters of the camera and the pixel coordinates representing the predicted position of the moving target to obtain the three-dimensional geographical coordinates corresponding to the predicted position. Correspondingly, based on the pixel coordinate values in the 3D panoramic monitoring screen, map the three-dimensional geographical coordinates corresponding to the predicted position back to the video screen pixel coordinates again, and then use cv2.rectangle or cv2.circle in OpenCV to draw an icon (i.e., the 3D virtual target model) on the frame and display it, so as to realize mapping the moving target into the 3D panoramic monitoring screen corresponding to the monitoring area and displaying it. For example, as Figure 5 shown, when a person (i.e., the moving target) within the red frame in the figure is detected, the person is displayed in the corresponding panoramic monitoring screen.

[0099] Among them, the 3D panoramic monitoring screen corresponding to the monitoring area is established after calibrating the installation height, the pitch angle, and the field of view of the camera according to the virtual human model and the virtual scale. Specifically, 3D modeling tools (such as Blender, 3ds Max) or laser scanning point cloud data can be used to create a 3D model of the monitoring area, including static elements such as buildings, roads, and obstacles, set a global 3D coordinate system (such as the east-north-up coordinate system), and map all cameras and target positions to this coordinate system. For example, the real 3D scene can be as Figure 6 shown. In this figure, the moving target is a 3D virtual human model, and the calibration of the 3D scene can be achieved by adjusting the values of the installation height, the pitch angle, and the field of view of the camera.

[0100] Step S103: Obtain a preset alarm rule, and perform real-time action tracking and detection on the moving target according to the preset alarm rule. If it is detected that the real-time action of the moving target meets the preset alarm rule, trigger the associated alarm response corresponding to the preset alarm rule.

[0101] Optionally, the alarm rule can be pre-configured, which is used to determine what specific situations require an alarm. In an optional embodiment of the present application, the preset alarm rule includes at least one of a target leaving post alarm rule, a sleeping on post alarm rule, an entering alarm rule, a leaving alarm rule, a target appearance alarm rule, a target disappearance alarm rule, a target classification alarm rule, a speed alarm rule, a direction alarm rule, an occlusion alarm rule, a tailing alarm rule, a flame alarm rule, a color alarm rule, an abandoned object alarm rule, an object moving alarm rule, and a traffic statistics alarm rule.

[0102] Optionally, the alarm rule can be set according to different moving targets. It can be set before the moving target is detected, or the alarm rule can be set after the moving target is determined. For example, as Figure 7 shown, after the moving target (i.e., the person outlined in red in the figure) is detected, the alarm rule is set.

[0103] Correspondingly, real-time action tracking and detection of the moving target can be performed according to the preset alarm rule. If the real-time action of the moving target meets the situation that requires an alarm in the alarm rule, trigger the associated alarm response corresponding to the preset alarm rule.

[0104] In an optional embodiment of the present application, triggering the associated alarm response corresponding to the preset alarm rule includes:

[0105] Determine at least one associated alarm device corresponding to the preset alarm rule, and send an alarm instruction to the at least one associated alarm device;

[0106] Record the accurate position and the corresponding real-time video stream of the moving target when the associated alarm response is triggered, and generate a record log by associating the accurate position and the corresponding real-time video stream of the moving target with the moving target management storage.

[0107] Optionally, different associated alarm devices can be set for different alarm rules. The specific type of the associated alarm device can be set by itself. For example, it can be a siren, a warning light, an access control device, a smoke sensor, an alarm host, an electronic fence, an infrared pair, etc. Correspondingly, when it is determined that the associated alarm response needs to be triggered, at least one associated alarm device corresponding to the alarm rule met this time can be determined, and then an alarm instruction is sent to the at least one associated alarm device. Correspondingly, after receiving the alarm instruction, the associated alarm device starts its corresponding alarm action, such as turning on the siren and warning light, closing the electronic fence, etc.

[0108] Optionally, in order to better know the real-time situation of this alarm afterwards, the accurate position corresponding to the moving target and the corresponding real-time video stream when triggering this linkage alarm response can be recorded at this time, and then the accurate position corresponding to the moving target and the corresponding real-time video stream are associated and stored with the moving target to generate a record log. Subsequently, the real-time situation of this alarm can be known based on this record log.

[0109] In the embodiment of the present application, for the monitoring area, the corresponding video stream can be obtained and the 3D panoramic monitoring corresponding to the monitoring area can be generated. Then, combined with the preset alarm rules, the real-time actions of the moving targets in the video stream are tracked and detected, and when the preset alarm rules are met, the linkage alarm response corresponding to the preset alarm rules is triggered. That is to say, in the present application, a panorama is generated by fusing multi-camera data, and early warning analysis is carried out by combining multi-dimensional data such as the target type, speed, and trajectory of the moving target with the alarm rules, realizing the early warning function of the system, changing the traditional video monitoring method that relies on human eyes to stare at the electronic eye and multi-camera rolling inspection of the pictures, and at the same time greatly improving the monitoring effectiveness, and being more suitable for complex security monitoring scenarios.

[0110] It can be understood that the intelligent video monitoring and alarm method provided in the embodiment of the present application can be executed by a high-definition intelligent video analysis server. As Figure 8 shown, this high-definition intelligent video analysis server can be linked with other devices to jointly realize the intelligent alarm of the monitoring area. For example, this high-definition intelligent video analysis server can form a topology network with a monitoring center, an intelligent early warning management platform server, a streaming media server, a storage server, an intelligent analysis alarm linkage dedicated controller, and a switch. The intelligent early warning management platform server can adjust and configure the intelligent early warning method for the monitoring area. The streaming media server can process the video stream of the monitoring area, and the storage server can store the video stream of the monitoring area. The high-definition intelligent video analysis server obtains the video stream captured by the front-end network camera through the switch, and realizes the alarm through linkage with other subsystems through the intelligent analysis alarm linkage dedicated controller when the alarm response is triggered.

[0111] Optionally, the method provided in the embodiment of the present application can be applied to various scenarios that require intelligent monitoring and alarm. For example, it can be divided into different applicable scenarios according to different functional types. Specifically, it can be as shown in the following table. Several scenarios are described below for exemplary illustration.

[0112]

[0113]

[0114] 1. Perimeter visualization intelligent early warning scenario

[0115] Such asFigure 9 As shown, the perimeter visualization intelligent early warning scenario applies the alarm method and system of intelligent video monitoring provided by the embodiments of the present application, changing the backward method of traditional video monitoring that relies on human eyes to stare at electronic eyes and multiple cameras to roll and inspect images, and is a transcendence and innovation of the existing video security monitoring system. Specifically, it can Figure 9 establish a 3D virtual scene based on the monitored area shown, configure multiple alarm rules (such as intrusion detection, tripwire detection, climbing over the wall, loitering detection, illegal parking and waste detection) for precise tracking and positioning. When it is found that the alarm rules are met, an alarm response can be triggered, such as controlling the audible and visual alarm (siren and warning light) and turning on the electronic fence and high-voltage power grid to realize the visual presentation of the alarm scene. At this time, it can not only play the role of video review, but also expand the perimeter prevention area and early warn of unsafe factors. It is currently widely used in important places such as prisons, government agencies, airports, ports, border defense, oil fields, schools, factories, warehouses, etc.

[0116] 2. Safe campus visualization intelligent early warning scenario

[0117] As Figure 10 shown, the safe campus visualization intelligent early warning scenario applies the alarm method and system of intelligent video monitoring provided by the embodiments of the present application, integrating campus high-definition video monitoring, perimeter prevention, face recognition, license plate recognition, one-key alarm, emergency command, campus fire protection, access control, broadcasting and other campus security prevention services, and is committed to protecting the personal safety of campus teachers and students, improving the campus security prevention system, and enhancing the overall campus prevention and control ability. The safe campus visualization scenario can detect the front-end working status, alarm records, alarm types, alarm events, and processing time after events based on the method provided by the present application, and perform unified data analysis and management. Therefore, campus management personnel at all levels can quickly understand the security operation situation, alarm handling situation, and characteristics of security events, providing a data basis for the management decision-making of school leaders.

[0118] 3. Visualization early warning scenario for work safety

[0119] As Figure 11As shown in the figure, the visualization early warning scenario for work safety applies the alarm method and system of intelligent video monitoring provided by the embodiments of the present application. At this time, the visualization intelligent early warning scenario for work safety integrates advanced technologies such as Internet Plus, Internet of Things, big data, cloud computing, and artificial intelligence, perfectly integrating intelligent monitoring and intelligent safety management. For example, it integrates power systems, fire protection systems, audio monitoring, perimeter alarms, access control systems, environmental systems, electronic patrols, and video monitoring, making work safety supervision and enterprise management more efficient, and can realize functions such as visualization of operation behavior safety, visualization of equipment monitoring at key positions, visualization of pre-event intelligent early warning, and visualization of system alarm linkage. It integrates work safety monitoring and control, intelligent control, intelligent scheduling, intelligent fire protection, real-time communication, and comprehensive management, and is applicable to the intelligent monitoring and management of work safety in various industries. For example, it is used for the emergency command, work safety monitoring, management scheduling, unattended operation, and other multi-faceted needs of government, military, large industry users, and enterprise users, achieving the purposes of intelligent monitoring, emergency command, alarm linkage, intelligent analysis, early warning prediction, and event recording, and realizing the upgrade, transformation, networking, unified management, and dispatching command of the system.

[0120] 4. Visualization intelligent early warning scenario for the civil explosive industry

[0121] As Figure 12 shown in the figure, since the civil explosive industry is a special industry with the dangerous attributes of flammability and explosiveness, work safety accidents often occur during production. Therefore, it is urgent to establish a more scientific, advanced, and reliable intelligent video monitoring system. At this time, the visualization intelligent early warning scenario for the civil explosive industry can be combined with the alarm method and system of intelligent video monitoring provided by the embodiments of the present application, integrating alarm linkage, off-duty detection, environmental control systems, intelligent analysis, perimeter alarms, and warehouse monitoring, improving the work safety management level of traditional high-risk manufacturing industries, enhancing the safety management level of the production site of civil explosive enterprises and the safety supervision level of the competent departments of the civil explosive industry, effectively preventing employees from violating regulations in operations and enterprises from producing illegally and irregularly, and providing effective means, real alarm linkage pictures, and effective data for the supervision department, providing a basis for supervision, and achieving pre-event prevention, efficient handling during the event, and post-event evidence collection.

[0122] 5. Visualization intelligent early warning scenario for off-duty and dozing

[0123] As Figure 13As shown in the figure, to meet the management needs and solve the supervision and management of the situation where staff leave their posts, the original manual inspection has been converted to automatic supervision, greatly reducing the labor intensity of the central management personnel and making the supervision more efficient. At this time, the alarm method and system of the intelligent video monitoring provided by the embodiments of the present application can be combined to implement an intelligent alarm system for leaving and sleeping on the post, realizing functions such as electronic supervision, video supervision, and intelligent government affairs. For the personnel leaving their posts and the leaving time, detailed records will be made for video storage, work discipline, performance appraisal, and service improvement. At the same time, the video is saved with traces. When the window personnel have doubts, they can view the monitoring video when leaving the post to retrieve and view the video, realizing post-event evidence collection.

[0124] 6. Intelligent Analysis of Big Data on Commercial Customer Behavior

[0125] As Figure 14 shown, the intelligent analysis system of big data on commercial customer behavior applies the alarm method and system of the intelligent video monitoring provided by the embodiments of the present application. At this time, based on video intelligent analysis, accurate pedestrian recognition and tracking can be carried out to achieve accurate passenger flow counting, and the passenger flow data is transmitted to the data analysis platform through the passenger flow analysis camera. The platform summarizes, sorts out, and analyzes according to the dimensions of the data, provides the data required by the user, conducts regional hot spot analysis, passenger flow statistical analysis, shelf attention analysis, consumption attribute analysis, and precision marketing, maximally excavates the sales potential of the shopping mall, increases profits, guides marketing activities with accurate data, maximally improves the activity effect, reasonably allocates resources, opens up sources on the one hand and saves money on the other hand, creates the best economic benefits, and effectively converts passenger flow data into economic benefits. It is mainly applied to large shopping malls, shopping centers, retail brand chain enterprises, stations, airports, various venues, and other places that require real-time reliable, continuous and accurate statistics of passenger flow and detection of passenger flow trends.

[0126] Optionally, the embodiments of the present application further provide an alarm system for intelligent video monitoring. This alarm system can be applied to an intelligent monitoring and alarm platform. As Figure 15 shown, this alarm system may include: a video acquisition module 501, a target detection and tracking module 502, a 3D mapping display module 503, and a behavior analysis and alarm module 504, where

[0127] The video acquisition module is used to support accessing cameras according to a set protocol and obtain the real-time video stream corresponding to the monitoring area through hardware decoding;

[0128] The target detection and tracking module: is used to detect the real-time video stream based on the target detection model, obtain the moving targets included in the real-time video stream, extract the motion data of the moving targets, and track the moving targets in real time according to the motion data of the moving targets;

[0129] 3D Mapping Display Module: It is used to obtain the installation parameters of the cameras corresponding to the monitoring area and the 3D virtual target model corresponding to the moving target, perform coordinate transformation on the pixel coordinates of the predicted position based on the installation parameters of the cameras to obtain the three-dimensional geographical coordinates corresponding to the predicted position, and map the 3D virtual target model into the pixel coordinates of the 3D panoramic monitoring screen and display it based on the three-dimensional geographical coordinates corresponding to the predicted position;

[0130] Behavior Analysis and Alarm Module: It is used to configure expandable preset alarm rules, perform real-time action tracking and detection on the moving target according to the preset alarm rules, and trigger a linkage alarm response and generate a record log when it is detected that the real-time action of the moving target meets the preset alarm rules.

[0131] In an optional embodiment of the present application, the alarm system is further configured with an SDK docking module. The SDK docking module includes an SDK development kit, preset interfaces, and example programs of the linkage subsystem. The SDK development kit includes a standard network encapsulation protocol, event stream operations, preset interfaces, and source code of the development document example program.

[0132] Optionally, the alarm system is further configured with an SDK docking module, which can provide a complete SDK development kit. The SDK development kit provides rich interfaces and practical example programs. Customers can conveniently perform subsequent development according to their own requirements, realize the integrated linkage with other subsystems, perform highly customized customization, and also include source code of the standard network protocol encapsulation for audio, video, and event stream operations / Metadata interface / HTTP API test tool development document example program.

[0133] Optionally, when the target detection and tracking module detects the real-time video stream based on the target detection model and obtains the moving target included in the real-time video stream, it is specifically used for:

[0134] Dynamically model the background of the real-time video stream based on the adaptive Gaussian mixture model to obtain the modeled real-time video stream;

[0135] Perform optical flow processing on the modeled real-time video stream to obtain the inter-frame motion vector field corresponding to the real-time video stream;

[0136] Detect the area of the inter-frame motion vector field based on the target detection model to obtain the moving target included in the real-time video stream.

[0137] Optionally, when the target detection and tracking module detects the area of the inter-frame motion vector field based on the target detection model and obtains the moving target included in the real-time video stream, it is specifically used for:

[0138] Detect the area of the inter-frame motion vector field based on the target detection model to obtain at least one initial moving target;

[0139] At least one initial moving target is filtered by using a multi - level combined filtering mechanism to obtain the moving targets included in the real - time video stream. The multi - level combined filtering mechanism includes at least one of the tracking type, target type, size threshold, and confidence threshold.

[0140] Optionally, the adaptive Gaussian mixture model is dynamically modeled based on the following formula:

[0141]

[0142] Where X t is the current pixel value, K is the number of Gaussian distributions, is the weight, u k,t and, ∑ k,t are the mean and covariance matrix α base is the base learning rate, ΔI t is the average pixel difference between the current frame and the background model, ΔS t is the average value of the optical flow amplitude between adjacent frames, I max is the normalization coefficient of the pixel difference, S max is the normalization coefficient of the optical flow amplitude, M k,t is the binary indicator function.

[0143] Optionally, when the target detection and tracking module tracks the moving target in real - time according to the motion data including position and speed of the moving target, it is specifically used for:

[0144] Based on the position and speed of the moving target, at least one predicted speed and at least one predicted position of the moving target are determined based on the Kalman filtering algorithm;

[0145] The trajectory of at least one predicted position of the moving target is smoothed to obtain at least one processed predicted position, and the predicted motion trajectory of the moving target is obtained according to at least one processed predicted position and at least one predicted speed;

[0146] The identifier of the moving target is determined, and the predicted motion trajectory of the moving target is associated with the identifier to achieve real - time tracking of the moving target.

[0147] Optionally, the predicted position is represented based on pixel coordinates. When the 3D mapping display module maps the moving target to the 3D panoramic monitoring screen corresponding to the monitoring area and displays it, it is specifically used for:

[0148] Obtain the installation parameters of the camera corresponding to the monitoring area and the 3D virtual target model corresponding to the moving target. The installation parameters include the installation height, pitch angle, and field of view;

[0149] Based on the installation parameters of the camera, the coordinate transformation of the pixel coordinates of the predicted position is performed to obtain the three - dimensional geographical coordinates corresponding to the predicted position;

[0150] Based on the three-dimensional geographical coordinates corresponding to the predicted position, map the 3D virtual target model into the pixel coordinates of the 3D panoramic surveillance screen and display it. The 3D panoramic surveillance screen is established after calibrating the installation height, pitch angle, and field of view of the camera according to the virtual human model and the virtual scale.

[0151] Optionally, the preset alarm rules include at least one of the target leaving the post alarm rule, sleeping on the post alarm rule, entering alarm rule, leaving alarm rule, target appearance alarm rule, target disappearance alarm rule, target classification alarm rule, speed alarm rule, direction alarm rule, occlusion alarm rule, tailing alarm rule, flame alarm rule, color alarm rule, abandoned object alarm rule, object moving alarm rule, and traffic statistics alarm rule.

[0152] Optionally, when the behavior analysis alarm module triggers the associated alarm response corresponding to the preset alarm rule, it is specifically used for:

[0153] Determine at least one associated alarm device corresponding to the preset alarm rule, and send an alarm instruction to at least one associated alarm device;

[0154] Record the accurate position corresponding to the moving target and the corresponding real-time video stream when the associated alarm response is triggered, and generate a record log by associating the accurate position corresponding to the moving target and the corresponding real-time video stream with the moving target management storage.

[0155] An alarm system for intelligent video surveillance according to this embodiment can execute an alarm method for intelligent video surveillance shown in the embodiments of the present application. The implementation principle is similar and will not be elaborated here.

[0156] The embodiments of the present application provide an electronic device. The electronic device in the embodiments of the present application includes: a processor; and a memory configured to store machine-readable instructions that, when executed by the processor, cause the processor to execute an alarm method for intelligent video surveillance.

[0157] The embodiments of the present application provide an electronic device, as Figure 16 shown, Figure 16 The electronic device shown includes a processor 2001 and a memory 2003. Among them, the processor 2001 and the memory 2003 are connected, such as through a bus 2002. Optionally, the electronic device 2000 may further include a transceiver 2004. It should be noted that in practical applications, the transceiver 2004 is not limited to one, and the structure of the electronic device 2000 does not constitute a limitation to the embodiments of the present application.

[0158] The processor 2001 can be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logic blocks, modules, and circuits described in connection with the disclosure of this application. The processor 2001 can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0159] The bus 2002 can include a path for transmitting information between the above components. The bus 2002 can be a PCI bus or an EISA bus, etc. The bus 2002 can be divided into an address bus, a data bus, a control bus, etc. For the sake of simplicity of representation, Figure 16 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0160] The memory 2003 can be a ROM or other types of static storage devices that can store static information and instructions, a RAM, or other types of dynamic storage devices that can store information and instructions. It can also be an EEPROM, a CD-ROM, or other optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic storage media, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0161] The memory 2003 is used to store the application program code for executing the solution of this application and is controlled by the processor 2001 to execute. The processor 2001 is used to execute the application program code stored in the memory 2003 to implement Figure 15 the actions of each device in an alarm system for intelligent video surveillance provided by the illustrated embodiment.

[0162] It should be understood that although the steps in the flowchart of the accompanying drawings are displayed sequentially in the direction of the arrows, these steps are not necessarily executed sequentially in the direction of the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings can include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same moment, but can be executed at different moments, and their execution order is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0163] The above are only some embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An alarm method for intelligent video surveillance, characterized in that, Including: Obtain the real-time video stream corresponding to the monitoring area and the target detection model, and detect the real-time video stream based on the target detection model to obtain the moving targets included in the real-time video stream; Extract the motion data of the moving targets, and track the moving targets in real time according to the motion data of the moving targets, and map the moving targets to the 3D panoramic monitoring screen corresponding to the monitoring area and display them; Obtain the preset alarm rule, and perform real-time action tracking detection on the moving targets according to the preset alarm rule. If it is detected that the real-time actions of the moving targets meet the preset alarm rule, trigger the associated alarm response corresponding to the preset alarm rule.

2. The method according to claim 1, wherein The detecting the real-time video stream based on the target detection model to obtain the moving targets included in the real-time video stream includes: Dynamically model the background of the real-time video stream based on the adaptive Gaussian mixture model to obtain the modeled real-time video stream; Perform optical flow processing on the modeled real-time video stream to obtain the inter-frame motion vector field corresponding to the real-time video stream; Detect the area of the inter-frame motion vector field based on the target detection model to obtain the moving targets included in the real-time video stream.

3. The method according to claim 2, characterized in that, The detecting the area of the inter-frame motion vector field based on the target detection model to obtain the moving targets included in the real-time video stream includes: Detect the area of the inter-frame motion vector field based on the target detection model to obtain at least one initial moving target; Adopt a multi-level joint filtering mechanism to filter the at least one initial moving target to obtain the moving targets included in the real-time video stream, and the multi-level joint filtering mechanism includes at least one of tracking type, target type, size threshold, and confidence threshold.

4. The method according to claim 2, wherein The adaptive Gaussian mixture model performs dynamic modeling based on the following formula: Among them, X t is the current pixel value, K is the number of Gaussian distributions, w k,t is the weight, u k,t and, ∑ k,t are the mean and covariance matrix α base is the base learning rate, ΔI t is the average pixel difference between the current frame and the background model, ΔS t is the mean value of the optical flow amplitude between adjacent frames, I max is the normalization coefficient of the pixel difference, S max is the normalization coefficient of the optical flow amplitude, M k,t is the binary indicator function.

5. The method according to claim 1, characterized in that, The motion data includes position and speed, and the tracking the moving targets in real time according to the motion data of the moving targets includes: Based on the position and speed of the moving target, determine at least one predicted speed and at least one predicted position of the moving target based on the Kalman filter algorithm; Perform trajectory smoothing processing on at least one predicted position of the moving target to obtain at least one processed predicted position, and obtain the predicted motion trajectory of the moving target according to the at least one processed predicted position and the at least one predicted speed; Determine the identifier of the moving target, and associate the predicted motion trajectory of the moving target with the identifier to achieve real-time tracking of the moving target.

6. The method according to claim 5, wherein The predicted position is represented based on pixel coordinates, and the mapping the moving target to the 3D panoramic monitoring screen corresponding to the monitoring area and displaying it includes: Obtain the installation parameters of the camera corresponding to the monitoring area and the 3D virtual target model corresponding to the moving target, and the installation parameters include installation height, pitch angle, and field of view; Perform coordinate transformation on the pixel coordinates of the predicted position based on the installation parameters of the camera to obtain the three-dimensional geographical coordinates corresponding to the predicted position; Based on the three-dimensional geographic coordinates corresponding to the predicted position, map the 3D virtual target model into the pixel coordinates of the 3D panoramic monitoring screen and display it. The 3D panoramic monitoring screen is established after calibrating the installation height, pitch angle, and field of view of the camera according to the virtual character model and the virtual scale.

7. The method according to claim 1, wherein The preset alarm rules include at least one of the target leaving post alarm rule, sleeping on post alarm rule, entering alarm rule, leaving alarm rule, target appearance alarm rule, target disappearance alarm rule, target classification alarm rule, speed alarm rule, direction alarm rule, occlusion alarm rule, trailing alarm rule, flame alarm rule, color alarm rule, abandoned object alarm rule, object moving alarm rule, and traffic statistics alarm rule.

8. The method according to claim 1, wherein Triggering the associated alarm response corresponding to the preset alarm rule includes: Determining at least one associated alarm device corresponding to the preset alarm rule and sending an alarm instruction to the at least one associated alarm device; Recording the accurate position and the corresponding real-time video stream of the moving target when the associated alarm response is triggered, and generating a record log by storing the accurate position and the corresponding real-time video stream of the moving target together with the moving target management.

9. An alarm system for intelligent video surveillance, characterized in that, The alarm system includes: A video acquisition module for supporting the access of cameras according to a set protocol and obtaining the real-time video stream corresponding to the monitoring area through hardware decoding; A target detection and tracking module: for detecting the real-time video stream based on a target detection model to obtain the moving targets included in the real-time video stream, extracting the motion data of the moving targets, and tracking the moving targets in real time according to the motion data of the moving targets; A 3D mapping display module: for obtaining the installation parameters of the camera corresponding to the monitoring area and the 3D virtual target model corresponding to the moving target, performing coordinate conversion on the pixel coordinates of the predicted position based on the installation parameters of the camera to obtain the three-dimensional geographic coordinates corresponding to the predicted position, and mapping the 3D virtual target model into the pixel coordinates of the 3D panoramic monitoring screen and displaying it based on the three-dimensional geographic coordinates corresponding to the predicted position; A behavior analysis and alarm module: for configuring expandable preset alarm rules, performing real-time action tracking detection on the moving targets according to the preset alarm rules, and triggering an associated alarm response and generating a record log when it is detected that the real-time actions of the moving targets meet the preset alarm rules.

10. The system according to claim 9, characterized in that, The alarm system is also configured with an SDK docking module. The SDK docking module includes an SDK development kit, preset interfaces, and example programs of the associated subsystems. The SDK development kit includes a standard network encapsulation protocol, event stream operations, preset interfaces, and source codes of example programs in the development documentation.

Images