Multi-camera linkage tracking control method and system

Abstract

The invention provides a multi-camera linkage tracking control method and system, and relates to the field of cameras, and the method comprises the steps: determining the prediction probability that a corresponding target object enters a visual blind area of each camera based on the predicted future movement track of each target object; selecting the visual blind area of which the prediction probability exceeds a preset probability threshold value, marking the selected visual blind area as a target visual blind area, determining all cameras of which the view field overlapping area contains the target visual blind area, and constructing a candidate camera set; when at least two cameras in the candidate camera set lock the same shooting area at the same time, selecting the camera with the highest shooting priority as a target camera; and controlling the target camera to adjust the shooting view angle. According to the method, the camera with the highest priority is selected, the subjectivity of randomly distributing the cameras is eliminated, shooting conflicts caused by the fact that the multiple cameras scramble for the same target object in the shooting area at the same time are avoided, it is kept that the spare cameras can track other target objects, and seamless relay of tracking of the multiple cameras is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of cameras, and in particular to a multi-camera linkage tracking control method and system. Background Art

[0002] Camera target tracking technology uses a camera to capture video sequences and utilizes computer vision algorithms to continuously locate and analyze the trajectory of target objects. It is widely used in security monitoring, autonomous driving, smart retail, indoor monitoring and other fields.

[0003] In related technologies, when tracking multiple objects simultaneously using multiple cameras, blind spots inevitably exist within the monitoring area due to the limited field of view of the cameras. This can easily cause the target object to escape the monitoring range of all cameras during movement. Furthermore, when multiple cameras need to switch perspectives to track a target object that enters a blind spot, or when the fields of view of multiple cameras overlap and the same object needs to be captured simultaneously, shooting conflicts are very likely to occur. In particular, when multiple cameras need to adjust their perspectives simultaneously to track multiple target objects, it is easy for multiple cameras to compete for the same target object within the shooting area, causing shooting conflicts and interrupting the tracking of other target objects. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a multi-camera linkage tracking control method and system to solve the problem of shooting conflict when multiple cameras track multiple target objects.

[0005] In a first aspect, the present application provides a multi-camera linkage tracking control method, comprising: Determine the predicted probability of the corresponding target object entering the visual blind spot of each camera based on the predicted future movement trajectory of each target object; Select a visual blind spot whose predicted probability exceeds a preset probability threshold and mark it as a target visual blind spot, determine all cameras whose field of view overlaps and includes the target visual blind spot, and construct a candidate camera set; When at least two cameras in the candidate camera set lock onto the same shooting area at the same time, the camera with the highest shooting priority is selected as the target camera; The target camera is controlled to adjust the shooting angle so that the shooting angle of the target camera includes the target visual blind spot.

[0006] In one embodiment, determining the predicted probability of the corresponding target object entering the visual blind spot of each camera based on the predicted future movement trajectory of each target object specifically includes: Discretizing the future movement trajectory into trajectory points arranged in time order; Calculating the shortest Euclidean distance from each trajectory point to the visual blind spot of each camera; The shortest Euclidean distance is used as a variable, and the predicted probability of the target object entering the corresponding visual blind spot is determined based on the Sigmoid function.

[0007] In one embodiment, before determining the predicted probability of the corresponding target object entering the visual blind spot of each camera based on the future movement trajectory, the method further includes: Construct a digital grid map of the monitoring area based on the real-time video streams, field of view parameters and location coordinates of multiple cameras; The overlapping areas of the fields of view and the blind areas of the cameras are determined in the digital grid map.

[0008] In one embodiment, determining the overlapping areas of the fields of view and blind spots of the cameras in the digital grid map specifically includes: Constructing a field of view spatial model of each camera according to the intrinsic parameter matrix, extrinsic parameter matrix and field of view range parameters of each camera, and determining the corresponding polygonal area in the geographic coordinate system according to the field of view spatial model; Calculate the field of view overlap of any two cameras; the field of view overlap is the ratio of the intersection area of ​​the polygonal areas corresponding to the two cameras to the sum of the polygonal areas of the two cameras; When the field of view overlap is greater than a preset overlap threshold, marking the intersection area of ​​the two cameras as a field of view overlap area in the digital grid map; The area in the digital grid map that is not covered by the field of view space model is marked as a visual blind spot of the corresponding camera.

[0009] In one embodiment, when at least two cameras in the candidate camera set simultaneously lock onto the same shooting area, selecting a camera with the highest shooting priority as the target camera specifically includes: Monitor the view switching instruction queues of each camera in real time. When the view switching instructions of at least two cameras involve the same overlapping area of ​​the field of view and have overlapping timestamps, determine that there is a shooting task conflict and mark the corresponding camera as a shooting conflict camera. The priority of each shooting conflict camera is calculated based on the target tracking task parameters of the shooting conflict camera based on the hierarchical analysis method, and the camera with the highest shooting priority is selected as the target camera.

[0010] In one embodiment, the calculating the priority of each conflicting camera based on the target tracking task parameters of the conflicting cameras based on the analytic hierarchy process specifically includes: Determining a judgment criterion corresponding to the target tracking task parameters; the target tracking task parameters include a tracking confidence value, a number of continuous tracking frames, and a preset priority level; the judgment criterion includes a confidence criterion, a trajectory continuity criterion, and a basic priority criterion; Based on the relative importance of each of the judgment criteria, determining a weight value of the corresponding judgment criterion and recording it as a first weight value; Based on the relative sizes of the target tracking task parameters of the corresponding shooting tasks in the shooting task conflict under each judgment criterion, a weight value of each shooting task under each judgment criterion is determined and recorded as a second weight value; the corresponding shooting tasks in the shooting task conflict include the currently executed first shooting task and the second shooting task for the same locked shooting area; The priority of each conflicting camera is determined according to the first weight value and the second weight value.

[0011] In one embodiment, determining a weight value of the corresponding judgment criterion based on the relative importance of each judgment criterion and recording it as a first weight value specifically includes: Comparing the importance of each criterion in the target tracking task parameters to determine the relative weight between any two criterion parameters; Constructing a criterion layer judgment matrix based on the relative weights; The criterion layer judgment matrix is ​​normalized, and the weight value of each criterion parameter in the criterion layer judgment matrix after the normalization is calculated by a summation method and recorded as a first weight value.

[0012] In one embodiment, determining the weight value of each shooting task under each judgment criterion and recording it as the second weight value based on the relative magnitudes of the target tracking task parameters corresponding to the shooting task in each judgment criterion in the shooting task conflict specifically includes: determining a ratio of target tracking task parameters corresponding to each judgment criterion between the first shooting task and the second shooting task; Construct a scheme-level judgment matrix based on the ratio corresponding to each judgment criterion; The solution layer judgment matrix is ​​normalized, and the weight value of each criterion parameter in the solution layer judgment matrix after normalization is calculated by a summation method and recorded as a second weight value.

[0013] In one embodiment, after constructing the criterion-level judgment matrix and / or the solution-level judgment matrix, the method further includes: Calculating the maximum eigenvalue of the criterion layer judgment matrix and / or the solution layer judgment matrix; determining a consistency index and an average random consistency index based on the maximum eigenvalue and the number of judgment criteria, and determining a consistency ratio based on a ratio of the consistency index to the average random consistency index; When the consistency ratio is lower than a preset ratio threshold, the criterion layer judgment matrix and / or the solution layer judgment matrix are reconstructed.

[0014] In a second aspect, the present application provides a multi-camera linkage tracking control system, comprising a processor and a memory; wherein the memory stores a computer program, and the computer program is used by the processor to load and execute the multi-camera linkage tracking control method as described in any one of the first aspects.

[0015] In the multi-camera linkage tracking control method and system of the present embodiment, the detection of the target visual blind spot and the construction of the corresponding candidate camera set are triggered only when the predicted probability of the target object entering the target visual blind spot exceeds the preset probability threshold. This can avoid the triggering probability of low-risk targets, reduce the probability of shooting conflict scenes, concentrate computing power on high-risk shooting conflict areas, and select the camera with the highest priority to shoot the high-risk shooting conflict area. This can eliminate the subjectivity of random camera allocation and avoid multiple cameras competing for the same target object in the shooting area at the same time to cause shooting conflicts. There are still free cameras to track other target objects, ensuring seamless relay of multiple camera tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 A flowchart of a multi-camera linkage tracking control method provided for one embodiment of the present application.

[0018] Figure 2 A schematic structural diagram of an electronic device provided in accordance with an embodiment of the present application. DETAILED DESCRIPTION

[0019] Specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the described embodiments are merely some, and not all, of the embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the description of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0020] In the description of the present invention, unless otherwise specified or limited, the terms "disposed," "installed," and "connected" should be understood broadly. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; and direct or indirect connections through an intermediary. Those skilled in the art will understand the specific meanings of these terms based on the specific circumstances.

[0021] The directions or positional relationships indicated by terms such as "upper", "lower", "left", "right", "front", "back", "top", "bottom", "inside" and "outside" are based on the directions or positional relationships shown in the accompanying drawings, or are the directions or positional relationships in which the inventive product is usually placed when in use. They are only for the convenience and simplification of description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0022] The terms "first," "second," "third," etc. are merely used to distinguish elements of similar nature and do not indicate or imply relative importance or a particular order.

[0023] The terms "comprises," "comprising," or any other variations thereof, are intended to cover a non-exclusive inclusion of elements other than the listed elements and may also include additional elements not specifically listed.

[0024] like Figure 1 As shown, this embodiment provides a multi-camera linkage tracking control method, including: Step S100: determining the predicted probability of the corresponding target object entering the visual blind spot of each camera based on the predicted future movement trajectory of each target object; Step S200: selecting a blind spot whose predicted probability exceeds a preset probability threshold and marking it as a target blind spot, determining all cameras whose overlapping fields of view include the target blind spot, and constructing a candidate camera set; Step S300: When at least two cameras in the candidate camera set lock onto the same shooting area at the same time, the camera with the highest shooting priority is selected as the target camera; Step S400: controlling the target camera to adjust the shooting angle so that the shooting angle of the target camera includes the target visual blind spot.

[0025] In the multi-camera linkage tracking control method of this embodiment, the detection of the target visual blind spot and the construction of the corresponding candidate camera set are triggered only when the predicted probability of the target object entering the target visual blind spot exceeds the preset probability threshold. This can avoid the triggering probability of low-risk targets, reduce the probability of shooting conflict scenes, concentrate computing power on high-risk shooting conflict areas, and select the camera with the highest priority to shoot the high-risk shooting conflict area. This can eliminate the subjectivity of random camera allocation and avoid multiple cameras competing for the same target object in the shooting area at the same time to cause shooting conflicts. There are still free cameras to track other target objects, ensuring seamless relay of multiple camera tracking.

[0026] Step S100: determining the predicted probability of the corresponding target object entering the visual blind spot of each camera based on the predicted future movement trajectory of each target object.

[0027] The future trajectory of a target object is a continuous sequence of spatial positions predicted by a target object motion model, which can be generated using a Kalman filter or a deep learning prediction algorithm. Exemplarily, the future trajectory includes a linear extrapolation trajectory based on the velocity vector or a nonlinear trajectory that takes into account steering behavior.

[0028] This embodiment preferably uses a Kalman filter to predict the future movement trajectory of the target object, and its state vector is ,in represents the coordinates of the target object, is the width of the target object, is the aspect ratio, and its expression is , is the height of the target object, Indicates the velocity component. In this embodiment, the aspect ratio Reconstructing the Kalman filter's state vector directly detects target distance through changes in width, resolving the issue of sudden size changes for small targets at long distances and ensuring more accurate prediction of the target's future trajectory. The state vector is obtained by iteratively updating observation data of the target's current state. The detailed process for predicting the target's future trajectory using the Kalman filter can be found in Chinese Patent Application No. 202410554844.4, "A Positioning Method and System Based on Time Difference of Arrival Information Using a Kalman Network."

[0029] Prediction probability quantifies the likelihood that a target's future trajectory will enter a specific blind spot. A mathematical model quantifies the spatial relationship between the target's trajectory and the blind spot as a value between 0 and 1. Its calculation relies on the Euclidean distance between the target's trajectory point and the blind spot boundary. This probabilistic prediction mechanism distinguishes between high-risk and low-risk areas, accurately screening targets for camera tracking changes and optimizing system resource allocation.

[0030] In one embodiment, the method of determining the predicted probability of the corresponding target object entering the visual blind spot of each camera based on the predicted future movement trajectory of each target object specifically includes: discretizing the future movement trajectory into trajectory points arranged in chronological order; calculating the shortest Euclidean distance from each trajectory point to the visual blind spot of each camera; and using the shortest Euclidean distance as a variable, determining the predicted probability of the target object entering the corresponding visual blind spot based on a Sigmoid function.

[0031] Trajectory points are discrete spatial coordinate sequences obtained by sampling at equal intervals. For example, trajectory points are generated at sampling intervals synchronized with the video frame rate, for example, 25 trajectory points are generated within the next second.

[0032] The shortest Euclidean distance is the minimum geometric distance between a trajectory point and the boundary of the visual blind spot, determined by a geometric calculation method. Exemplarily, the shortest Euclidean distance includes the minimum distance between a trajectory point and a vertex or edge of the blind spot boundary of a convex polygon.

[0033] When calculating the shortest Euclidean distance from each trajectory point to each camera's blind spot, the blind spot is modeled as a convex polygon. The polygon vertex set is traversed and the distance between the trajectory point and each vertex and boundary is calculated, with the minimum value selected as the final distance. If the trajectory point is within the blind spot, the final distance value is marked as zero; otherwise, the minimum distance value is used. Through geometric space mapping, the topological relationship between the trajectory and the blind spot is converted into numerical features, providing input parameters for subsequent probabilistic modeling, thus avoiding the lack of scene adaptability caused by binary judgment in traditional methods.

[0034] Using the shortest Euclidean distance as a variable, the predicted probability of the target object entering the corresponding visual blind spot is determined based on the Sigmoid function. The expression of the Sigmoid function is: ,in For the The predicted probability of each trajectory point; is the slope factor, the default value is 5; is the buffer radius, with a default value of 1.5 meters. When the predicted probability of three consecutive trajectory points is greater than or equal to 0.7, it indicates that the target object has entered a high-risk area. At this point, it is necessary to determine the visual blind spot that the target object is about to enter and activate nearby cameras for tracking. By mapping continuous distance values ​​into a probability space through a nonlinear transformation, it is possible to distinguish the progressive risk level of the target object as it approaches the visual blind spot.

[0035] In addition, to ensure the real-time performance of trajectory prediction, trajectory prediction only calculates the trajectory points of 50 frames within the next 2 seconds and limits the number of blind spot polygon vertices to less than or equal to 8 to control computational complexity. At the same time, to prevent the target object from temporarily turning when approaching the visual blind spot, causing the trajectory to change, when the velocity vector angle of the target object's temporary turn is greater than 45°, the cumulative number of predicted probabilities is reset and the re-prediction of the future movement trajectory is triggered. It can accurately assess the risk level of the target entering the visual blind spot in complex scenarios while maintaining the system's real-time and anti-interference capabilities, improving its adaptability to dynamic scenarios such as the target object wandering at the edge of the blind spot, and providing a more refined risk quantification basis for subsequent camera linkage decision-making.

[0036] In one embodiment, before determining the predicted probability of the corresponding target object entering the visual blind spot of each camera based on the future movement trajectory, the method further includes: constructing a digital grid map of the monitoring area based on the real-time video streams, field of view range parameters and position coordinates of multiple cameras; and determining the field of view overlapping area and visual blind spot of each camera in the digital grid map.

[0037] Real-time video stream is a data stream of image sequences collected by the camera in real time. It is used to dynamically reflect the actual coverage of the monitoring area and can be obtained through the camera's network transmission interface or local storage.

[0038] The field of view parameters refer to the optical characteristics of the camera, such as the horizontal or vertical viewing angle and focal length, and can be obtained through the device factory calibration or user configuration file.

[0039] The location coordinates are the physical installation coordinates of the camera in three-dimensional space and can be obtained through GPS positioning, manual measurement, or indoor positioning systems.

[0040] A digital grid map is a spatial model that discretizes physical space into regular or irregular grid cells. Each cell stores its spatial coordinates and the identification of the cameras covering that area. When constructing a digital grid map, each camera's field of view is first projected onto the grid map using geometric projection, marking the grid cells it theoretically covers. Then, real-time video streams are used to analyze the camera's actual visible area, correcting for field of view deviations caused by obstructions or environmental changes. Finally, a uniform grid (such as equal-sized rectangular or hexagonal cells) or an adaptive grid (dynamically adjusting cell size based on camera distribution density) is used to integrate coverage information from all cameras to form a global coverage map.

[0041] The overlapped field of view is the set of grid cells covered by two or more cameras simultaneously, and can be identified by counting the number of cameras covering each grid cell. The blind spot is the set of grid cells not covered by any camera, also determined based on the number of cameras covering each grid cell.

[0042] In one embodiment, determining the overlapping areas of the fields of view and blind spots of the cameras in the digital grid map specifically includes: A field of view space model of each camera is constructed according to the intrinsic parameter matrix, extrinsic parameter matrix and field of view range parameters of each camera, and a polygonal area in the corresponding geographic coordinate system is determined according to the field of view space model.

[0043] The intrinsic parameter matrix is ​​a set of parameters that describe the optical characteristics of the camera, including focal length, image principal point coordinates, and distortion coefficients. It can be obtained through camera calibration experiments or factory parameters.

[0044] The extrinsic parameter matrix is ​​a set of parameters that defines the spatial relationship between the camera coordinate system and the world coordinate system. It consists of a rotation matrix R and a translation vector T and can be obtained through three-dimensional spatial coordinate measurement or real-time acquisition of sensor data.

[0045] The field of view parameters are the camera's horizontal field of view angle α and vertical field of view angle β. and Calculated, where and are the pixel width and height of the imaging plane, is the focal length.

[0046] The field of view spatial model is a geometric model that abstracts the camera's field of view into a three-dimensional cone, whose vertices are composed of the geographic coordinates of the camera's optical center and the boundary points determined by the field of view angle. For example, a near plane and a far plane are selected in the camera's cone model. The near plane is a plane 1 meter away from the camera's optical center, and the far plane is a plane corresponding to the camera's maximum monitoring distance (usually 30-50 meters). The coordinates of the boundary points of the near and far planes in the camera coordinate system are determined, and the coordinates of the boundary points are converted to geographic coordinates using external parameters. The boundary points of the near plane are connected in clockwise order to form a near plane polygon, and the boundary points of the far plane are connected in clockwise order to form a far plane polygon. The corresponding vertices of the near plane polygon and the far plane polygon are connected to form a closed polygonal area.

[0047] Calculate the field of view overlap of any two cameras; the field of view overlap is the ratio of the intersection of the polygonal areas corresponding to the two cameras to the sum of the areas of the polygonal areas of the two cameras. When the field of view overlap is greater than a preset overlap threshold, mark the intersection of the two cameras as a field of view overlap area in the digital grid map.

[0048] Field of view overlap is a metric that quantifies the degree of overlap between the fields of view of two cameras. It is calculated as the ratio of the intersection area of ​​two polygonal regions to the sum of the areas of the two regions. For example, for any two cameras A and B, the fields of view of both cameras A and B are converted into a 3D grid model in a geographic coordinate system. Each voxel in the grid is traversed to determine whether it belongs to the fields of view of both A and B. The total number of overlapping voxels and the total number of voxels in the fields of view of A and B are counted. The ratio of the total number of overlapping voxels to the total number of voxels in the fields of view of A and B is the field of view overlap of cameras A and B.

[0049] The overlap threshold is a set baseline value. When the field of view overlap exceeds the preset overlap threshold, the corresponding area is determined to be an overlapping area of ​​the two cameras' fields of view. For example, the polygon intersection area can be calculated using a scanline algorithm or triangle decomposition method. When the ratio exceeds a preset threshold (e.g., 0.3), the corresponding area is marked as a field of view overlap area.

[0050] The area in the digital grid map that is not covered by the field of view space model is marked as a visual blind spot of the corresponding camera.

[0051] Visual blind spots are areas within a digital grid map that are not covered by any camera's field of view spatial model. These blind spots are determined in conjunction with a real-time updated field of view model. For example, when a camera fails due to a malfunction or occlusion, the system removes that camera's coverage area from the overall coverage, and the newly uncovered area becomes a new blind spot. Furthermore, by calculating the straight-line distance between the blind spot center and each camera and the feasibility of viewing angle adjustment, a mapping relationship between blind spots and potential covering cameras can be established, providing a basis for subsequent viewing angle adjustments.

[0052] Furthermore, to adapt to dynamic environmental changes, a dynamic topology update mechanism is implemented to periodically collect camera internal and external parameters and refresh the field of view spatial model. For example, for mobile cameras, the system recollects parameters and updates the model every 100 milliseconds. When a camera goes offline or field of view overlap decreases, the system simultaneously adjusts the overlap area markers and blind spot ranges, triggering viewing angle adjustment instructions for the backup camera. This real-time spatial modeling and topological analysis overcomes the inability of traditional static models to adapt to environmental changes, improving system robustness.

[0053] By combining geometric modeling with dynamic analysis, the accuracy of field of view overlap and blind spot recognition is significantly improved, reducing the risk of misjudgment due to environmental changes. This provides reliable spatial data support for the coordinated tracking of multi-camera systems, effectively reducing the risk of tracking failure due to environmental changes or parameter errors, and improving the decision-making reliability and real-time performance of multi-camera systems.

[0054] Step S200: selecting a visual blind spot whose predicted probability exceeds a preset probability threshold and marking it as a target visual blind spot, determining all cameras whose field of view overlapping areas include the target visual blind spot, and constructing a candidate camera set.

[0055] The preset probability threshold is a system-defined trigger condition used to screen for high-risk visual blind spots. For example, if the predicted probability of three consecutive trajectory points exceeds 0.7, the target is considered to have entered a high-risk area.

[0056] The candidate camera set is a collection of all cameras that can cover the target's blind spot. This set must ensure that each camera's field of view at least partially covers the target blind spot. Geometric calculations can be used to determine whether each camera's field of view covers the blind spot. For example, a polygon containment algorithm can be used to verify the spatial relationship between the camera's field of view boundary and the blind spot's vertices. A dynamic triggering mechanism reduces computing resources in low-risk scenarios while ensuring timely resource allocation for high-risk scenarios. By screening the candidate camera set to form a pool of selectable monitoring resources, interference from low-probability events can be reduced, lowering computational complexity.

[0057] Step S300: When at least two cameras in the candidate camera set lock onto the same shooting area at the same time, the camera with the highest shooting priority is selected as the target camera.

[0058] The shooting priority is a weighted value calculated based on factors such as the camera's current load, distance to the target, and viewing angle adjustment cost. For example, a camera that is closer to the target may receive a higher priority. The highest priority score can be selected for competitive arbitration by comparing them. For example, when two cameras have the same priority, the idle time or hardware performance parameters of the cameras can be further compared. When it is detected that there are multiple cameras in the candidate camera set pointing to the same area at the same time, the system starts the priority evaluation mechanism, calculates the priority score for each camera, selects the camera with the highest score as the target camera, and the remaining cameras are excluded or reassigned. This is achieved by marking the target camera as the responsible device and notifying other cameras to release control of the corresponding area. This reduces resource competition through quantitative decision-making and ensures maximum system resource utilization.

[0059] In one embodiment, when at least two cameras in the candidate camera set simultaneously lock onto the same shooting area, selecting a camera with the highest shooting priority as the target camera specifically includes: Step S301: monitor the view switching instruction queue of each camera in real time. When the view switching instructions of at least two cameras involve the same field of view overlapping area and the timestamps overlap, determine that there is a shooting task conflict and mark the corresponding camera as a shooting conflict camera.

[0060] The view switching command queue is a structured data set used to record camera view adjustment requests. It is obtained by the camera receiving and storing view adjustment commands. For example, the view switching command queue includes fields such as a command identifier, a camera identifier, geographic boundary parameters, a timestamp parameter, and a status identifier (e.g., "pending," "in progress," or "completed").

[0061] Conflict detection is based on a set of rules based on both spatial and temporal constraints. For example, spatial overlap is determined by calculating whether the intersection area or volume of the geographic boundary parameters of two instructions exceeds a preset threshold, while temporal overlap is determined by comparing whether the timestamp intervals of the instructions intersect.

[0062] The tag of the conflicting camera is a camera that identifies the conflicting camera. Exemplarily, the tag is generated by the conflict detection module after comparing the candidate instructions in the instruction queue with each other.

[0063] Step S302: Calculate the priority of each conflicting camera based on the target tracking task parameters of the conflicting cameras based on the hierarchical analysis method, and select the camera with the highest priority as the target camera.

[0064] The priority calculation of the hierarchical analysis method is a mathematical method that constructs a multi-dimensional weight matrix and calculates the comprehensive score. The weight coefficient of each parameter is defined, and then the parameter value of each camera is standardized. Finally, the weighted sum is obtained to determine the priority.

[0065] In one embodiment, the calculating the priority of each conflicting camera based on the target tracking task parameters of the conflicting cameras based on the analytic hierarchy process specifically includes: Step S3021: Determine the judgment criteria corresponding to the target tracking task parameters; the target tracking task parameters include tracking confidence value, continuous tracking frame number and preset priority level; the judgment criteria include confidence criterion, trajectory continuity criterion and basic priority criterion.

[0066] The tracking confidence value is a quantitative indicator that characterizes the reliability of the target detection result. It corresponds to the confidence criterion and can be obtained from the probability value or confidence score output by the target detection algorithm. For example, the tracking confidence value includes a value between 0 and 1 based on the output of the deep learning model.

[0067] The continuous tracking frame count is a parameter that characterizes the continuity of a target's trajectory. It corresponds to the trajectory continuity criterion and is calculated by counting the duration of target tracking in consecutive video frames. For example, the continuous tracking frame count includes the cumulative number of frames in which the target is not lost within a preset time period.

[0068] The preset priority levels are predefined weight parameters based on task type or monitoring requirements, corresponding to the basic priority criteria, and are set through system configuration files or user input. For example, the preset priority levels include giving high-risk target tracking tasks a higher priority than normal target tracking tasks.

[0069] The confidence criterion is an evaluation standard for assessing the reliability of target detection, the trajectory continuity criterion is an important basis for measuring tracking stability, and the basic priority criterion is a priori condition reflecting the task type or monitoring requirement level. The three together constitute a hierarchical model of judgment criteria.

[0070] The priority of conflicting tasks is used as the target layer, the confidence criterion, trajectory continuity criterion and basic priority criterion are used as the criterion layer, and each conflicting task is used as the solution layer. The hierarchical model provides structural support for weight allocation by decomposing the comprehensive priority target into the criterion layer and the solution layer.

[0071] In one embodiment, the criteria layer parameters include tracking confidence, trajectory continuity score, and pre-set priority levels. The solution layer corresponds to multiple conflicting camera tasks. For example, if a camera is tracking a high-risk target but has low confidence, the influence of various parameters can be balanced through weighting to prevent a single metric from dominating the decision.

[0072] Step S3022: Based on the relative importance of each of the judgment criteria, determine the weight value of the corresponding judgment criterion and record it as the first weight value.

[0073] In one embodiment, based on the relative importance of each of the judgment criteria, the weight value of the corresponding judgment criterion is determined and recorded as the first weight value, specifically including: comparing the importance of each judgment criterion in the target tracking task parameters to determine the relative weight between any two criterion parameters; constructing a criterion layer judgment matrix based on the relative weight; normalizing the criterion layer judgment matrix, and calculating the weight value of each criterion parameter in the normalized criterion layer judgment matrix by the summation method and recording it as the first weight value.

[0074] Importance comparison is a comparative process that assesses the relative importance of different judgment criteria. This is achieved by pairwise comparison of criterion parameters. The relative importance of target tracking task parameters corresponding to each criterion is determined based on expert experience or experimental data. For example, when tracking confidence has a greater impact on overall priority than trajectory continuity score, the ratio between the two is set to 3:1; when trajectory continuity score has a greater impact on overall priority than preset priority, the ratio is set to 2:1; and when preset priority has a greater impact on overall priority than tracking confidence, the ratio is set to 4:1. Based on these comparison results, a judgment matrix is ​​constructed, and the matrix element values ​​reflect the relative importance of each criterion.

[0075] The criterion-level judgment matrix is ​​a two-dimensional matrix structure used to numerically represent the relative weights of the various criteria. Its element values ​​are assigned using a 1-9 scale. For example, if two criteria are equally important, the corresponding element value is 1; if one criterion is slightly more important than the other, the corresponding element value is 3; if it is extremely important, the corresponding element value is 9. The construction of this matrix follows the rule of reciprocal values. For example, if the ratio of tracking confidence to trajectory continuity is 3:1, the corresponding element value is 3, and the inverse position is set to 1 / 3. This method creates a 3×3 matrix that comprehensively reflects the relative importance of the various criteria.

[0076] Normalization is the process of standardizing the column elements of the judgment matrix. This involves summing the elements of each column to obtain a column sum vector, and then dividing each element by the corresponding column sum. For example, the column sum of the elements [3, 1, 2] is 6. After normalization, the column elements become [0.5, 0.1667, 0.3333]. Calculating weight values ​​through the summation method specifically involves taking the arithmetic mean of the elements of the normalized matrix row by row, ultimately obtaining the weight values ​​of each criterion parameter. This process ensures that the sum of the weights is 1. For example, if the weights of tracking confidence, trajectory continuity, and preset priority are 0.4, 0.3, and 0.3, respectively, the sum is 1.

[0077] Furthermore, the rationality of weight assignment is verified by calculating the consistency ratio (CR). When CR ≤ 0.1, the weight is valid; otherwise, the judgment matrix elements must be adjusted and recalculated. This structured comparison process avoids arbitrary subjective weight assignments. Normalization ensures the mathematical rationality of weight assignments, while consistency testing effectively reduces priority deviations caused by weight inconsistencies. Ultimately, this provides scientific and interpretable weight foundational data support for multi-camera task scheduling.

[0078] Step S3023: Based on the relative sizes of the target tracking task parameters of the corresponding shooting tasks in each judgment criterion in the shooting task conflict, determine the weight value of each shooting task under each judgment criterion and record it as the second weight value; the corresponding shooting tasks in the shooting task conflict include the currently executed first shooting task and the second shooting task for the same locked shooting area.

[0079] In one embodiment, the weight value of each shooting task under each judgment criterion is determined based on the relative size of the target tracking task parameters of the corresponding shooting tasks in each judgment criterion in the shooting task conflict and recorded as the second weight value, specifically including: determining the ratio of the target tracking task parameters corresponding to each judgment criterion of the first shooting task and the second shooting task; constructing a scheme layer judgment matrix based on the ratio corresponding to each judgment criterion; normalizing the scheme layer judgment matrix, and calculating the weight value of each criterion parameter in the normalized scheme layer judgment matrix by the summation method and recording it as the second weight value.

[0080] The target tracking task parameter ratio is a numerical ratio used to quantify the relative advantages of different shooting tasks under the same judgment criteria. It is obtained by comparing the differences between the two tasks on specific criteria parameters and mapping them to a preset ratio rule. Exemplary ratio rules include, but are not limited to, mapping parameter differences to ratios under the tracking confidence criterion, discretizing frame ratios and ratios under the trajectory continuity scoring criterion, and nonlinear relationships between level differences and ratios under the preset priority criterion. For example, if task A is the first shooting task currently being executed, and task B is the second shooting task for the same locked shooting area, in the tracking confidence criterion, if the confidence of task A is 0.4 higher than that of task B, then the ratio is determined to be 4:1 using linear interpolation or preset rules based on the intermediate value of the difference between 0.3 and 0.5; in the trajectory continuity scoring criterion, if the number of continuous tracking frames of task A is 2.5 times that of task B, the ratio can be taken as 2:1 or 3:1 according to the system's preset discretization strategy; in the preset priority criterion, if the level of task A is 1.5 levels higher than that of task B, the ratio can be determined by linear interpolation or rounding up rules.

[0081] The scenario-level judgment matrix is ​​a two-dimensional matrix used to structure the ratio relationship between different shooting tasks. Its element values ​​are determined by the ratio of each task under a specific criterion. For example, if the ratio of task A to task B under the tracking confidence criterion is 2:1, the element value at the corresponding position in the matrix is ​​set to 2, and the reverse position is set to 0.5. It is understood that this matrix can be expanded to scenarios containing three or more tasks, and the relative weight calculation between multiple tasks can be realized by increasing the dimension.

[0082] The specific process of normalizing the scheme layer judgment matrix and calculating the weight value of each criterion parameter in the normalized scheme layer judgment matrix by the summation method is similar to the above-mentioned process of generating the first weight value and will not be repeated here.

[0083] Step S3024: Determine the priority of each conflicting camera according to the first weight value and the second weight value.

[0084] The final priority is determined through a weighted calculation of multiple criteria. The weights at the criterion level (the first weight value) and the solution level (the second weight value) are weighted and summed to obtain a comprehensive priority score for each camera task. For example, in a security scenario, if two cameras are simultaneously tracking the same high-risk target but have significantly different confidence levels, the system prioritizes the camera with the higher confidence level based on weights, ensuring reliable tracking of the critical target.

[0085] In steps S3021-S3024, a hierarchical model is constructed to decompose the priority evaluation objectives, a judgment matrix is ​​used to quantify the relative importance of parameters and verify the rationality of the weights, and a dynamic adjustment mechanism is combined to adapt to real-time task status changes. Finally, a comprehensive priority score is calculated through multi-criteria weighted calculation to achieve a balance between multi-dimensional requirements such as task type, detection reliability, and tracking stability, reduce the risk of tracking interruption caused by subjective or static strategies, and significantly improve the technical effect of system resource utilization and task completion quality in complex scenarios.

[0086] In one embodiment, after constructing the criterion layer judgment matrix and / or the scheme layer judgment matrix, it also includes: calculating the maximum eigenvalue of the criterion layer judgment matrix and / or the scheme layer judgment matrix; determining the consistency index and the average random consistency index based on the maximum eigenvalue and the number of judgment criteria, and determining the consistency ratio based on the ratio of the consistency index to the average random consistency index; when the consistency ratio is lower than a preset ratio threshold, reconstructing the criterion layer judgment matrix and / or the scheme layer judgment matrix.

[0087] Maximum eigenvalue is the largest eigenvalue in the judgment matrix and can be obtained by eigenvalue decomposition or numerical calculation algorithms. For example, for a 3×3 criterion layer judgment matrix, its maximum eigenvalue can be estimated by numerical methods such as power iteration method.

[0088] The consistency index CI is a quantitative parameter that measures the degree to which the judgment matrix deviates from complete consistency. Its calculation formula is: , where n is the number of criteria.

[0089] The average random consistency index (RI) is the average consistency index of the randomly generated judgment matrix based on the matrix dimension statistics. For example, its value is obtained by table lookup, for example, when the matrix dimension is 3, it corresponds to 0.58, and when the dimension is 4, it corresponds to 0.90.

[0090] The consistency ratio is the standardized consistency test result, which is the ratio of the consistency index to the average random consistency index. For example, when the consistency index is 0.05 and the average random consistency index is 0.58, the consistency ratio is approximately 8.6%. The preset ratio threshold is the critical value for determining whether the judgment matrix is ​​acceptable. For example, its value can be set to 0.1 or lower depending on the application scenario requirements.

[0091] When the consistency ratio is lower than the preset threshold, matrix reconstruction is triggered. The judgment matrix element values ​​can be adjusted through local correction or global reset strategies. For example, the ratio of contradictory element pairs can be adjusted from 4:1 to 3:1, or the pairwise comparison steps can be re-executed to eliminate logical contradictions, thereby ensuring the reliability of weight calculation.

[0092] By forcing the consistency ratio to be lower than a threshold, logical contradictions in the judgment matrix can be effectively identified and corrected, preventing the weight distribution from deviating from actual needs. At the same time, through strategies such as dynamic threshold adjustment, local correction, or global reset, adaptive optimization of the judgment matrix can be achieved, ensuring that the weight distribution is consistent with both domain experience and mathematical consistency requirements, ultimately improving the rationality of multi-camera system resource allocation and the success rate of tracking tasks.

[0093] In steps S301-S304, structured instruction queue management is used to ensure the comprehensiveness of conflict detection, and dual thresholds are used to filter non-critical conflicts to reduce false alarms. At the same time, multi-dimensional priority decisions are converted into computable quantitative indicators to avoid the subjectivity and delay problems of manual intervention. This can achieve orderly allocation of camera resources and seamless connection of continuous tracking in complex scenarios.

[0094] Step S400: Control the target camera to adjust the shooting angle so that the shooting angle of the target camera includes the target visual blind spot.

[0095] Viewing angle adjustment refers to the process of changing the observation area of ​​a camera by adjusting the focal length, rotation angle, or translation mechanism. The parameter adjustment must meet the physical constraints of mechanical motion.

[0096] Enabling the target camera's field of view to include the target's blind spot means adjusting the camera's field of view so that it covers the target's blind spot. This adjustment requires that the camera's field of view at least include key areas within the blind spot, such as the center point or boundary vertices of the blind spot's convex polygon. The visible area can be altered by adjusting camera parameters such as the pan / tilt angle and focal length.

[0097] By sending control commands to the target camera, it drives its mechanical structure or adjusts electronic parameters to shift its viewing angle toward the target's blind spot. During this adjustment process, real-time feedback on the adjusted viewing angle is provided, and spatial coordinates are used to verify complete coverage of the target area. If not, further fine-tuning or a secondary assessment is triggered. This precise control ensures efficient utilization of camera resources, reducing the risk of the target escaping surveillance. Meanwhile, other cameras can continue tracking other targets, achieving seamless tracking tasks.

[0098] In summary, the multi-camera linkage tracking control method of this embodiment significantly improves the efficiency and reliability of multi-camera linkage tracking by screening high-risk blind spots based on predicted probability, constructing a candidate camera set to limit resource scope, resolving multi-camera conflicts through priority scoring, and ensuring blind spot coverage through dynamic viewing angle adjustment. This solution reduces conflict probability and optimizes resource utilization through a hierarchical decision-making mechanism. It also possesses dynamic adaptability and computing power optimization capabilities, making it suitable for complex scenarios requiring high-precision multi-target tracking.

[0099] Based on the same inventive concept as the above embodiment, this embodiment further provides a system, including a processor and a memory; wherein the memory stores a computer program, and the computer program is used to be loaded by the processor and execute the above method.

[0100] like Figure 2 As shown, based on the same inventive concept as the above embodiment, this embodiment further provides a computer-readable storage medium, which stores instructions for loading and executing the above method by a processor.

[0101] In the embodiments of the mobile terminal and computer-readable storage medium provided in this application, all technical features of the above-mentioned control method embodiments are included. The expanded and explained contents of the specification are basically the same as those of the above-mentioned method embodiments and will not be repeated here.

[0102] An embodiment of the present application further provides a computer program product, which includes computer program code. When the computer program code runs on a computer, the computer executes the methods in the various possible implementation modes described above.

[0103] An embodiment of the present application also provides a chip, including a memory and a processor, wherein the memory is used to store computer programs, and the processor is used to call and run the computer programs from the memory, so that a device equipped with the chip executes the methods in the various possible implementation modes as described above.

[0104] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0105] In this application, the same or similar terminology, technical solutions and / or application scenario descriptions are generally only described in detail the first time they appear. When they appear again later, they are generally not repeated for the sake of brevity. When understanding the technical solutions and other contents of this application, for the same or similar terminology, technical solutions and / or application scenario descriptions that are not described in detail later, you can refer to the previous relevant detailed descriptions.

[0106] In this application, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0107] The various technical features of the technical solution of this application can be combined arbitrarily. In order to make the description concise, not all possible combinations of the various technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0108] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product is stored in a storage medium as above, including a number of instructions for enabling a terminal device to execute the method of each embodiment of the present application. The above is only a preferred embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the content of the present application specification and drawings, or directly or indirectly used in other related technical fields, is similarly included in the scope of patent protection of the present application. It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other.

[0109] The foregoing description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be readily conceived by a person skilled in the art within the technical scope disclosed herein are intended to be encompassed within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A multi-camera linkage tracking control method, characterized in that: include: Determine the predicted probability of the corresponding target object entering the visual blind spot of each camera based on the predicted future movement trajectory of each target object; Select a visual blind spot whose predicted probability exceeds a preset probability threshold and mark it as a target visual blind spot, determine all cameras whose field of view overlaps and includes the target visual blind spot, and construct a candidate camera set; When at least two cameras in the candidate camera set lock onto the same shooting area at the same time, the camera with the highest shooting priority is selected as the target camera; The target camera is controlled to adjust the shooting angle so that the shooting angle of the target camera includes the target visual blind spot.

2. The multi-camera linkage tracking control method according to claim 1, characterized in that: The step of determining the predicted probability of the corresponding target object entering the visual blind spot of each camera based on the predicted future movement trajectory of each target object specifically includes: Discretizing the future movement trajectory into trajectory points arranged in time order; Calculating the shortest Euclidean distance from each trajectory point to the visual blind spot of each camera; The shortest Euclidean distance is used as a variable, and the predicted probability of the target object entering the corresponding visual blind spot is determined based on the Sigmoid function.

3. The multi-camera linkage tracking control method according to claim 1, characterized in that: Before determining the predicted probability of the corresponding target object entering the visual blind spot of each camera based on the future movement trajectory, the method further includes: Construct a digital grid map of the monitoring area based on real-time video streams, field of view parameters and location coordinates of multiple cameras; The overlapping areas of the fields of view and the blind areas of the cameras are determined in the digital grid map.

4. The multi-camera linkage tracking control method according to claim 3, characterized in that: Determining the overlapping areas of the fields of view and blind spots of the cameras in the digital grid map specifically includes: Constructing a field of view spatial model of each camera according to the intrinsic parameter matrix, extrinsic parameter matrix and field of view range parameters of each camera, and determining the corresponding polygonal area in the geographic coordinate system according to the field of view spatial model; Calculate the field of view overlap of any two cameras; the field of view overlap is the ratio of the intersection area of ​​the polygonal areas corresponding to the two cameras to the sum of the polygonal areas of the two cameras; When the field of view overlap is greater than a preset overlap threshold, marking the intersection area of ​​the two cameras as a field of view overlap area in the digital grid map; The area in the digital grid map that is not covered by the field of view space model is marked as a visual blind spot of the corresponding camera.

5. The multi-camera linkage tracking control method according to any one of claims 1 to 4, characterized in that: When at least two cameras in the candidate camera set simultaneously lock onto the same shooting area, selecting a camera with the highest shooting priority as the target camera specifically includes: Monitor the view switching instruction queues of each camera in real time. When the view switching instructions of at least two cameras involve the same overlapping area of ​​the field of view and have overlapping timestamps, determine that there is a shooting task conflict and mark the corresponding camera as a shooting conflict camera. The priority of each shooting conflict camera is calculated based on the target tracking task parameters of the shooting conflict camera based on the hierarchical analysis method, and the camera with the highest shooting priority is selected as the target camera.

6. The multi-camera linkage tracking control method according to claim 5, characterized in that: The step of calculating the priority of each conflicting camera based on the target tracking task parameters of the conflicting cameras based on the analytic hierarchy process specifically includes: Determining a judgment criterion corresponding to the target tracking task parameters; the target tracking task parameters include a tracking confidence value, a continuous tracking frame number, and a preset priority level; the judgment criterion includes a confidence criterion, a trajectory continuity criterion, and a basic priority criterion; Based on the relative importance of each of the judgment criteria, determining a weight value of the corresponding judgment criterion and recording it as a first weight value; Based on the relative sizes of the target tracking task parameters of the corresponding shooting tasks in the shooting task conflict under each judgment criterion, a weight value of each shooting task under each judgment criterion is determined and recorded as a second weight value; the corresponding shooting tasks in the shooting task conflict include the currently executed first shooting task and the second shooting task for the same locked shooting area; The priority of each conflicting camera is determined according to the first weight value and the second weight value.

7. The multi-camera linkage tracking control method according to claim 6, characterized in that: The step of determining a weight value of a corresponding judgment criterion based on the relative importance of each judgment criterion and recording the weight value as a first weight value specifically includes: Comparing the importance of each criterion in the target tracking task parameters to determine the relative weight between any two criterion parameters; Constructing a criterion layer judgment matrix based on the relative weights; The criterion layer judgment matrix is ​​normalized, and the weight value of each criterion parameter in the criterion layer judgment matrix after the normalization is calculated by a summation method and recorded as a first weight value.

8. The multi-camera linkage tracking control method according to claim 6, characterized in that: The method of determining a weight value of each shooting task under each judgment criterion based on the relative size of the target tracking task parameters corresponding to the shooting task in each judgment criterion in the shooting task conflict and recording the weight value as the second weight value specifically includes: determining a ratio of target tracking task parameters corresponding to each judgment criterion between the first shooting task and the second shooting task; Construct a scheme-level judgment matrix based on the ratio corresponding to each judgment criterion; The solution layer judgment matrix is ​​normalized, and the weight value of each criterion parameter in the solution layer judgment matrix after normalization is calculated by a summation method and recorded as a second weight value.

9. The multi-camera linkage tracking control method according to claim 7 or 8, characterized in that: After constructing the criterion-level judgment matrix and / or the solution-level judgment matrix, the following steps are also included: Calculating the maximum eigenvalue of the criterion layer judgment matrix and / or the solution layer judgment matrix; determining a consistency index and an average random consistency index based on the maximum eigenvalue and the number of judgment criteria, and determining a consistency ratio based on a ratio of the consistency index to the average random consistency index; When the consistency ratio is lower than a preset ratio threshold, the criterion layer judgment matrix and / or the solution layer judgment matrix are reconstructed.

10. A multi-camera linkage tracking control system, comprising a processor and a memory; wherein, The memory stores a computer program, which is used for the processor to load and execute the multi-camera linkage tracking control method as described in any one of claims 1 to 9.

Images