Self-adaptive multi-mode switching image displacement monitoring method and system

Abstract

The invention relates to the technical field of image processing and displacement monitoring, in particular to a self-adaptive multi-mode switching image displacement monitoring method and system, which can dynamically match different tracking strategies by resolving the displacement speed, the acceleration, the track smoothness and the shielding rate of each target in real time, so that under the situation of a complex tracking scene, the tracking accuracy is improved. A tracking strategy with higher occupation of computing resources and better resolving performance can be scheduled; in the case of a simple tracking scene, a tracking strategy with lower occupation of computing resources and higher response speed can be scheduled, and the problems of tracking precision reduction and even target loss caused by inapplicability of a single algorithm are effectively avoided, so that the accuracy and efficiency of displacement monitoring are improved. Meanwhile, by monitoring the total number of tracking targets and the utilization rate of computer resources in real time, intelligent thread scheduling is carried out on a tracking strategy determined for each target, and the problems that the processing frame rate is reduced and the system is blocked due to a traditional single thread scheduling mechanism are solved.

Description

Technical Field

[0001] This invention relates to the field of image processing and displacement monitoring technology, and in particular to an adaptive multi-mode switching image displacement monitoring method and system. Background Technology

[0002] Image displacement monitoring technology is used in many engineering fields. Existing image displacement monitoring systems typically rely on pre-set, fixed-type target tracking algorithms to calculate target displacement. These algorithms can exhibit good tracking accuracy and stability under specific ideal conditions. However, in real-world engineering applications, factors such as the target's motion amplitude and speed, the frequency and duration of target occlusion, and the available computing resources all exhibit significant dynamic changes and uncertainties. Using a single, fixed tracking strategy is difficult to adapt to complex and changing tracking conditions, leading to a sharp decline in tracking accuracy, or even target loss, thus affecting the reliability of monitoring.

[0003] Specifically, in near-infrared image monitoring applications, when multiple targets need to be tracked and their displacement calculated simultaneously, using the traditional single-threaded serial processing mechanism to track and calculate each target one by one will lead to a decrease in the processing frame rate, causing system lag and resulting in the loss of tracked targets. On the other hand, in monitoring scenarios where the target displacement is small and the target shape is stable, using complex tracking algorithms will increase the consumption of computing resources, introduce unnecessary computational delays, and ultimately reduce the overall tracking performance and efficiency of the system.

[0004] In summary, the field of image displacement monitoring technology lacks the ability to adaptively adjust based on real-time operating conditions and system status. Therefore, there is an urgent need in this field for an adaptive image displacement monitoring method to adapt to the complex and ever-changing practical displacement monitoring application requirements and achieve stable, efficient, and flexible displacement detection. Summary of the Invention

[0005] The present invention aims to provide an adaptive multi-mode switching image displacement monitoring method and system to adapt to the complex and ever-changing actual displacement monitoring application requirements and achieve stable, efficient and flexible displacement detection.

[0006] To achieve the above objectives, the first aspect of the present invention provides an image displacement monitoring method with adaptive multi-mode switching, comprising the following steps: Acquire the current frame image and a set of historical consecutive frame images, and identify several tracking targets in the set of historical consecutive frame images; Obtain the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate for each tracked target; Based on the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target, determine the current tracking strategy corresponding to each tracked target; Obtain the total number of tracked targets and computer resource utilization; Based on the total number of tracking targets and the computer resource utilization rate, thread scheduling is performed on the current tracking strategy corresponding to each tracking target, and then the current coordinates of each tracking target in the current frame image are calculated through the current tracking strategy corresponding to each tracking target. Based on the historical continuous frame image set, the coordinates of the previous frame of each tracked target are obtained. Then, based on the current coordinates and the coordinates of the previous frame of each tracked target, the relative displacement of each tracked target is calculated to complete the image displacement monitoring.

[0007] The aforementioned adaptive multi-mode switching image displacement monitoring method calculates the displacement velocity, acceleration, trajectory smoothness, and occlusion rate of each target in real time, and dynamically matches different tracking strategies based on these parameters. This allows for the scheduling of tracking strategies with higher computational resource consumption and better computational performance in complex tracking scenarios, and tracking strategies with lower computational resource consumption and faster response speed in simple tracking scenarios. This effectively avoids the problem of decreased tracking accuracy or even target loss due to the inapplicability of a single algorithm, thereby improving the accuracy and efficiency of displacement monitoring.

[0008] Meanwhile, this invention intelligently schedules threads for each target's tracking strategy by real-time monitoring of the total number of tracked targets and computer resource utilization. This prevents excessive computer resource utilization from causing computational lag or system failure, thus overcoming the frame rate drop and system lag problems caused by traditional single-thread scheduling mechanisms. Therefore, the aforementioned adaptive multi-mode switching image displacement monitoring method of this invention adapts to the complex and ever-changing practical displacement monitoring application requirements by adaptively switching between the current tracking strategy and thread scheduling method. It can stably, efficiently, and flexibly meet the complex and ever-changing practical engineering application needs.

[0009] Further, obtaining the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target includes: Based on the historical continuous frame image set, a first historical frame image and a second historical frame image are obtained, and then the inter-frame time difference between the first historical frame image and the second historical frame image is obtained; For any of the aforementioned tracking targets: Obtain the first coordinates of the tracked target in the first historical frame image, and obtain the second coordinates of the tracked target in the second historical frame image; The current displacement velocity is obtained based on the first coordinate, the second coordinate, and the inter-frame time difference; The historical displacement velocity of the tracked target is obtained based on the historical continuous frame image set, and then the displacement acceleration is obtained based on the current displacement velocity and the historical displacement velocity.

[0010] In this implementation, the current displacement velocity is calculated based on the changes in the tracked target's coordinates and the time difference between adjacent historical frames. Furthermore, the displacement acceleration is calculated by comparing the current velocity with historical velocities. By effectively characterizing the instantaneous intensity and complexity of the tracked target's motion using displacement velocity and acceleration, it becomes easier to switch the appropriate tracking strategy based on the complexity of the target's motion. For scenarios with complex target motion, a tracking strategy with higher computational resource consumption and better computational performance can be scheduled; for scenarios with simple target motion, a tracking strategy with lower computational resource consumption and faster response speed can be scheduled. Ultimately, this improves the adaptability of displacement monitoring, enhancing the accuracy and robustness of target tracking while maintaining system response speed.

[0011] Further, obtaining the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target includes: For any of the aforementioned tracking targets: Based on the historical continuous frame image set, obtain several historical coordinates of the tracked target; Historical displacement trajectories are obtained based on several historical coordinates, the curvature of the historical displacement trajectories is calculated, and the smoothness of the trajectory is obtained based on the curvature. Based on the historical continuous frame image set, obtain several historical target images of the tracked target; Obtain the number of historical pixels corresponding to each historical target image, and then, based on the number of historical pixels corresponding to each historical target image and a preset occlusion ratio threshold, select a number of occluded target images from the number of historical target images. The occlusion rate is obtained by acquiring the proportion of all occluded target images in all historical target images.

[0012] In this implementation, the smoothness of the trajectory is evaluated by calculating the curvature of the historical displacement trajectory. This effectively characterizes the intensity of the target's motion and distinguishes between a smooth, stable motion and an irregular motion such as jitter or maneuvering. Thus, when the target is in a stable motion, a tracking strategy with higher computational resource consumption and better solution performance can be scheduled; when the target is in a violent motion, a tracking strategy with lower computational resource consumption and faster response speed can be scheduled. Ultimately, the tracking accuracy is improved while ensuring the system's response speed.

[0013] Furthermore, for the number of historical pixels corresponding to historical target images, when the number of historical pixels is lower than a preset effective pixel ratio threshold, the tracked target in that historical target image frame is considered to be occluded. The occlusion rate is obtained by statistically analyzing the proportion of occluded target images in historical target images, thus achieving an objective quantification of the frequency and degree of target occlusion. This parameter enables the system to identify high-risk tracking targets in complex scenes with frequent occlusion and to activate tracking strategies with stronger predictive capabilities, enhancing the system's reliability in complex real-world scenarios. Therefore, this implementation method, through a dual evaluation of target motion quality and scene complexity, achieves refined and intelligent selection of tracking strategies, further ensuring the accuracy and stability of displacement monitoring results.

[0014] Further, determining the current tracking strategy for each tracking target based on its current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate includes: For any of the aforementioned tracking targets: If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate satisfy the first preset condition, then a preset tracker is set as the current tracking strategy; If the occlusion rate meets the second preset condition, then the CSRT tracker is set as the preset tracker; otherwise, the KCF tracker is set as the preset tracker.

[0015] If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate do not meet the first preset condition, then a preset motion prediction model is set as the current tracking strategy.

[0016] In this implementation, a hierarchical and refined tracking strategy selection and switching mechanism is constructed. When the target's motion state meets the first preset condition, i.e., complex situations such as high speed, high acceleration, or severe trajectory jitter occur, it is determined that high-precision real-time tracking is required to ensure the accuracy of displacement calculation. Therefore, a preset tracker with higher computational resource consumption but superior performance is activated for precise locking. Based on this, the specific tracker selected is determined according to the occlusion rate. For targets with a high occlusion rate, the CSRT tracker with stronger anti-interference capabilities is selected to ensure tracking stability; for targets with a low occlusion rate, the KCF tracker with a faster response speed is selected to moderately optimize efficiency while ensuring accuracy.

[0017] When the target's motion state does not meet the first preset condition, namely, a stable state with low speed and smooth trajectory, the system determines that the tracked target is in a stable state. At this time, the system no longer uses the computationally intensive tracker, but switches to a lightweight preset motion prediction model, directly calculating the current coordinates based on a set of historical continuous frames. This reduces the consumption of computational resources and significantly improves the system's response speed and processing frame rate. Therefore, this implementation achieves optimal allocation of limited computational resources, ensuring that the image displacement monitoring method maintains efficient and stable operation in various scenarios.

[0018] Further, the step of scheduling threads for the current tracking strategy corresponding to each tracking target based on the total number of tracking targets and the computer resource utilization rate includes: If the total number of tracked targets and the computer resource utilization rate meet the third preset condition, then an independent thread is allocated to each current tracking strategy so that all current tracking strategies can run in parallel; If the total number of tracked targets and the computer resource utilization rate do not meet the third preset condition, then all the current tracking strategies will run in a preset execution priority order.

[0019] In this implementation, an adaptive thread scheduling mechanism based on system computer resource load is introduced. When the total number of tracked targets is large and the computer resource utilization rate is low, meeting the third preset condition, the system determines that it has the capability for parallel processing. At this time, the system allocates an independent computing thread for the tracking strategy of each target, enabling it to run in parallel, shortening the total time required to process all tracked targets, significantly improving the efficiency of multi-target concurrent processing, and thus ensuring stable, efficient, and flexible displacement detection in multi-target tracking scenarios.

[0020] When the total number of tracked targets is small or computer resource utilization is high, and the third preset condition is not met, the system switches to a single-threaded sequential execution mode. In this mode, all tracking strategies run sequentially according to preset execution priorities, effectively avoiding the additional system overhead caused by creating and managing multiple threads in low-load or resource-constrained scenarios, and preventing system lag or instability caused by resource contention. Therefore, this implementation allows the system to dynamically and intelligently switch between single-threaded and multi-threaded execution modes based on real-time operating conditions, ensuring both processing efficiency in complex scenarios and system stability and resource conservation in simple scenarios or when resources are scarce.

[0021] Further, the step of calculating the current coordinates of each tracking target in the current frame image using the current tracking strategy corresponding to each tracking target includes: For any of the tracking targets, if the current tracking strategy for that tracking target is a preset tracker: The preset tracker is initialized based on the tracking target; The current frame image and the set of historical consecutive frame images are input to the preset tracker, and then the current coordinates of the tracked target in the current frame image are output through the preset tracker. For any of the tracking targets, if the current tracking strategy corresponding to the tracking target is a preset motion prediction model: The coordinates of the previous frame of the tracked target are obtained based on the historical continuous frame image set; According to the Kalman filter algorithm, a state transition matrix is ​​constructed based on the current displacement velocity, the displacement acceleration and the coordinates of the previous frame corresponding to the tracked target, and then the motion prediction model is constructed based on the state transition matrix; The current coordinates of the tracked target in the current frame image are obtained based on the motion prediction model.

[0022] In this implementation, for the strategy employing a preset tracker, the preset tracker, as an algorithm module, has the core function of accurately locating the target by analyzing its visual features within consecutive image frames. By inputting the current frame image and a set of historical consecutive frames into the initialized tracker, the system can dynamically match and lock onto the target using its appearance information. This method, which directly relies on image pixel information, enables the system to achieve accurate positioning even when the target's motion is complex, its appearance has changed partially, or there is strong background interference, thus ensuring the displacement monitoring accuracy for multi-target tracking in complex tracking scenarios.

[0023] On the other hand, for strategies employing pre-defined motion prediction models, this implementation avoids complex analysis by the tracker, thus saving computer resources. In this implementation, based on the Kalman filter algorithm, a state transition matrix is ​​constructed using only the target's previous frame coordinates, displacement velocity, and acceleration, quickly calculating the current coordinates. This avoids the computational overhead of tracker image processing, achieving rapid prediction of the target's position and significantly improving the system's target displacement monitoring response speed and operational efficiency in simple tracking scenarios.

[0024] Therefore, this implementation method adapts to the complex and ever-changing actual displacement monitoring application requirements by configuring differentiated execution mechanisms for different strategies, and ultimately achieves stable, efficient and flexible displacement detection.

[0025] A second aspect of the present invention provides an adaptive multi-mode switching image displacement monitoring system, comprising: The frame image processing module is used to acquire the current frame image and a set of historical continuous frame images, and to identify several tracking targets in the set of historical continuous frame images; The target tracking analysis module is used to obtain the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target; The tracking strategy switching module is used to determine the current tracking strategy for each tracking target based on the current displacement velocity, displacement acceleration, trajectory smoothness and occlusion rate of each tracking target. The current coordinate calculation module is used to obtain the total number of tracking targets and the computer resource utilization rate. Based on the total number of tracking targets and the computer resource utilization rate, it performs thread scheduling for the current tracking strategy corresponding to each tracking target, and then calculates the current coordinates of each tracking target in the current frame image through the current tracking strategy corresponding to each tracking target. The displacement calculation module is used to obtain the previous frame coordinates of each tracked target based on the historical continuous frame image set, and then calculate the relative displacement of each tracked target based on the current coordinates and the previous frame coordinates of each tracked target, thereby completing image displacement monitoring.

[0026] Further, obtaining the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target includes: Based on the historical continuous frame image set, a first historical frame image and a second historical frame image are obtained, and then the inter-frame time difference between the first historical frame image and the second historical frame image is obtained; For any of the aforementioned tracking targets: Obtain the first coordinates of the tracked target in the first historical frame image, and obtain the second coordinates of the tracked target in the second historical frame image; The current displacement velocity is obtained based on the first coordinate, the second coordinate, and the inter-frame time difference; The historical displacement velocity of the tracked target is obtained based on the historical continuous frame image set, and then the displacement acceleration is obtained based on the current displacement velocity and the historical displacement velocity.

[0027] Further, obtaining the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target includes: For any of the aforementioned tracking targets: Based on the historical continuous frame image set, obtain several historical coordinates of the tracked target; Historical displacement trajectories are obtained based on several historical coordinates, the curvature of the historical displacement trajectories is calculated, and the smoothness of the trajectory is obtained based on the curvature. Based on the historical continuous frame image set, obtain several historical target images of the tracked target; Obtain the number of historical pixels corresponding to each historical target image, and then, based on the number of historical pixels corresponding to each historical target image and a preset occlusion ratio threshold, select a number of occluded target images from the number of historical target images. The occlusion rate is obtained by acquiring the proportion of all occluded target images in all historical target images.

[0028] Further, determining the current tracking strategy for each tracking target based on its current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate includes: For any of the aforementioned tracking targets: If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate satisfy the first preset condition, then a preset tracker is set as the current tracking strategy; If the occlusion rate meets the second preset condition, then the CSRT tracker is set as the preset tracker; otherwise, the KCF tracker is set as the preset tracker.

[0029] If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate do not meet the first preset condition, then a preset motion prediction model is set as the current tracking strategy. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating an adaptive multi-mode switching image displacement monitoring method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an adaptive multi-mode switching image displacement monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0031] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that the following detailed descriptions are exemplary and intended to provide further detailed explanation of the invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein in the specification is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings are used to distinguish different objects, not to describe a particular order.

[0032] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0033] The present invention aims to provide an adaptive multi-mode switching image displacement monitoring method and system to adapt to the complex and ever-changing actual displacement monitoring application requirements and achieve stable, efficient and flexible displacement detection.

[0034] Please refer to Figure 1 To achieve the above objectives, the first aspect of the present invention provides an image displacement monitoring method with adaptive multi-mode switching, comprising the following steps: S101. Obtain the current frame image and the set of historical continuous frame images, and identify several tracking targets in the set of historical continuous frame images; S102. Obtain the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target; S103. Based on the current displacement velocity, displacement acceleration, trajectory smoothness and occlusion rate of each of the tracked targets, determine the current tracking strategy corresponding to each of the tracked targets; S104. Obtain the total number of tracked targets and computer resource utilization; S105. Based on the total number of tracking targets and the computer resource utilization rate, perform thread scheduling on the current tracking strategy corresponding to each tracking target, and then calculate the current coordinates of each tracking target in the current frame image through the current tracking strategy corresponding to each tracking target. S106. Based on the historical continuous frame image set, obtain the previous frame coordinates of each of the tracked targets, and then calculate the relative displacement of each of the tracked targets based on the current coordinates and the previous frame coordinates of each of the tracked targets to complete image displacement monitoring.

[0035] It should be noted that in the historical continuous frame image set, the frame image that is closest to the current frame image on the time axis is taken as the previous frame image, and the coordinates of the tracking target in the previous frame image are the coordinates of the tracking target in the previous frame.

[0036] The aforementioned adaptive multi-mode switching image displacement monitoring method calculates the displacement velocity, acceleration, trajectory smoothness, and occlusion rate of each target in real time, and dynamically matches different tracking strategies based on these parameters. This allows for the scheduling of tracking strategies with higher computational resource consumption and better computational performance in complex tracking scenarios, and tracking strategies with lower computational resource consumption and faster response speed in simple tracking scenarios. This effectively avoids the problem of decreased tracking accuracy or even target loss due to the inapplicability of a single algorithm, thereby improving the accuracy and efficiency of displacement monitoring.

[0037] Meanwhile, this invention intelligently schedules threads for each target's tracking strategy by real-time monitoring of the total number of tracked targets and computer resource utilization. This prevents excessive computer resource utilization from causing computational lag or system failure, thus overcoming the frame rate drop and system lag problems caused by traditional single-thread scheduling mechanisms. Therefore, the aforementioned adaptive multi-mode switching image displacement monitoring method of this invention adapts to the complex and ever-changing practical displacement monitoring application requirements by adaptively switching between the current tracking strategy and thread scheduling method. It can stably, efficiently, and flexibly meet the complex and ever-changing practical engineering application needs.

[0038] In a preferred embodiment, step S101 specifically involves the image acquisition and initial target calibration process.

[0039] In this embodiment, during the system initialization phase, the user is guided through a graphical user interface (GUI) to manually select each near-infrared light source target as the tracking target. The system stores the initial position selected by the user as a reference template for each target and synchronously records the initial frame position, unique ID number, and current frame number of each target, thus completing the system initialization.

[0040] During the actual displacement monitoring phase, a near-infrared camera was used for image acquisition, and the acquired target image frames were grayscale images.

[0041] Further, step S102, which involves obtaining the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target, includes: Based on the historical continuous frame image set, a first historical frame image and a second historical frame image are obtained, and then the inter-frame time difference between the first historical frame image and the second historical frame image is obtained; For any of the aforementioned tracking targets: Obtain the first coordinates of the tracked target in the first historical frame image, and obtain the second coordinates of the tracked target in the second historical frame image; The current displacement velocity is obtained based on the first coordinate, the second coordinate, and the inter-frame time difference; The historical displacement velocity of the tracked target is obtained based on the historical continuous frame image set, and then the displacement acceleration is obtained based on the current displacement velocity and the historical displacement velocity.

[0042] In one specific embodiment, the first historical frame image and the second historical frame image are the two frame images in the historical continuous frame image set that are closest to the current frame image on the time axis.

[0043] In this implementation, the current displacement velocity is calculated based on the changes in the tracked target's coordinates and the time difference between adjacent historical frames. Furthermore, the displacement acceleration is calculated by comparing the current velocity with historical velocities. By effectively characterizing the instantaneous intensity and complexity of the tracked target's motion using displacement velocity and acceleration, it becomes easier to switch the appropriate tracking strategy based on the complexity of the target's motion. For scenarios with complex target motion, a tracking strategy with higher computational resource consumption and better computational performance can be scheduled; for scenarios with simple target motion, a tracking strategy with lower computational resource consumption and faster response speed can be scheduled. Ultimately, this improves the adaptability of displacement monitoring, enhancing the accuracy and robustness of target tracking while maintaining system response speed.

[0044] Further, step S102, which involves obtaining the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target, includes: For any of the aforementioned tracking targets: Based on the historical continuous frame image set, obtain several historical coordinates of the tracked target; Historical displacement trajectories are obtained based on several historical coordinates, the curvature of the historical displacement trajectories is calculated, and the smoothness of the trajectory is obtained based on the curvature. Based on the historical continuous frame image set, obtain several historical target images of the tracked target; Obtain the number of historical pixels corresponding to each historical target image, and then, based on the number of historical pixels corresponding to each historical target image and a preset occlusion ratio threshold, select a number of occluded target images from the number of historical target images. The occlusion rate is obtained by acquiring the proportion of all occluded target images in all historical target images.

[0045] In this implementation, the smoothness of the trajectory is evaluated by calculating the curvature of the historical displacement trajectory. This effectively characterizes the intensity of the target's motion and distinguishes between a smooth, stable motion and an irregular motion such as jitter or maneuvering. Thus, when the target is in a stable motion, a tracking strategy with higher computational resource consumption and better solution performance can be scheduled; when the target is in a violent motion, a tracking strategy with lower computational resource consumption and faster response speed can be scheduled. Ultimately, the tracking accuracy is improved while ensuring the system's response speed.

[0046] Furthermore, for the number of historical pixels corresponding to historical target images, when the number of historical pixels is lower than a preset effective pixel ratio threshold, the tracked target in that historical target image frame is considered to be occluded. The occlusion rate is obtained by statistically analyzing the proportion of occluded target images in historical target images, thus achieving an objective quantification of the frequency and degree of target occlusion. This parameter enables the system to identify high-risk tracking targets in complex scenes with frequent occlusion and to activate tracking strategies with stronger predictive capabilities, enhancing the system's reliability in complex real-world scenarios. Therefore, this implementation method, through a dual evaluation of target motion quality and scene complexity, achieves refined and intelligent selection of tracking strategies, further ensuring the accuracy and stability of displacement monitoring results.

[0047] Further, step S103, which determines the current tracking strategy for each tracking target based on the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracking target, includes: For any of the aforementioned tracking targets: If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate satisfy the first preset condition, then a preset tracker is set as the current tracking strategy; If the occlusion rate meets the second preset condition, then the CSRT tracker is set as the preset tracker; otherwise, the KCF tracker is set as the preset tracker.

[0048] If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate do not meet the first preset condition, then a preset motion prediction model is set as the current tracking strategy.

[0049] In this implementation, a hierarchical and refined tracking strategy selection and switching mechanism is constructed. When the target's motion state meets the first preset condition, i.e., complex situations such as high speed, high acceleration, or severe trajectory jitter occur, it is determined that high-precision real-time tracking is required to ensure the accuracy of displacement calculation. Therefore, a preset tracker with higher computational resource consumption but superior performance is activated for precise locking. Based on this, the specific tracker selected is determined according to the occlusion rate. For targets with a high occlusion rate, the CSRT tracker with stronger anti-interference capabilities is selected to ensure tracking stability; for targets with a low occlusion rate, the KCF tracker with a faster response speed is selected to moderately optimize efficiency while ensuring accuracy.

[0050] When the target's motion state does not meet the first preset condition, namely, a stable state with low speed and smooth trajectory, the system determines that the tracked target is in a stable state. At this time, the system no longer uses the computationally intensive tracker, but switches to a lightweight preset motion prediction model, directly calculating the current coordinates based on a set of historical continuous frames. This reduces the consumption of computational resources and significantly improves the system's response speed and processing frame rate. Therefore, this implementation achieves optimal allocation of limited computational resources, ensuring that the image displacement monitoring method maintains efficient and stable operation in various scenarios.

[0051] In a preferred embodiment, the adaptive multi-mode switching selection of the current tracking strategy is completed through steps S102 and S103. This mechanism not only supports both manual setting and automatic decision-making, but also achieves refined and adaptive management of the tracking strategy through comprehensive evaluation of multi-dimensional information.

[0052] In this embodiment, the system first performs a global condition assessment to determine whether to enable the tracker. Specifically, when the total number of tracked targets reaches a preset threshold N (e.g., 3) or when occlusion is frequently observed through statistical analysis of the effective pixel ratio of the targets, the system will determine it as a complex tracking scenario and enable the preset tracker to ensure the stability of target tracking. Conversely, if the system analysis shows that the average motion speed and acceleration of all targets are low and the trajectory is smooth, indicating a simple scenario, the tracker can be turned off, and position prediction can be performed solely using the preset motion prediction model based on the information from the previous frame to save computer resources.

[0053] In addition, the system will monitor the current frame rate and CPU / GPU resource usage in real time. When resources are scarce, it will prioritize shutting down trackers for some targets to ensure the real-time performance of core functions.

[0054] After deciding to enable the tracker, the system will further select a strategy for each target. For targets in complex scenes, such as those with significant changes in lighting or frequent occlusion, the system selects the more powerful CSRT tracker as the current tracking strategy for the corresponding target; while for relatively simple scenes or situations with limited system resources, the more computationally efficient KCF tracker is selected to improve running speed.

[0055] In a preferred embodiment, the system continuously monitors the tracker's performance during operation. For example, if the frame rate fluctuation of the CSRT tracker exceeds a preset stability threshold within a short period, the system automatically downgrades it to a KCF tracker to ensure real-time performance. Once the frame rate stabilizes, the system can switch back to the CSRT tracker, thus achieving a dynamic balance between tracking accuracy and operational efficiency. Therefore, this preferred embodiment, by integrating individual target states and the system's global load for comprehensive decision-making, enables the entire displacement monitoring system to proactively adapt to complex and ever-changing application environments.

[0056] Further, step S105, which involves thread scheduling for the current tracking strategy corresponding to each tracking target based on the total number of tracking targets and the computer resource utilization rate, includes: If the total number of tracked targets and the computer resource utilization rate meet the third preset condition, then an independent thread is allocated to each current tracking strategy so that all current tracking strategies can run in parallel; If the total number of tracked targets and the computer resource utilization rate do not meet the third preset condition, then all the current tracking strategies will run in a preset execution priority order.

[0057] In this implementation, an adaptive thread scheduling mechanism based on system computer resource load is introduced. When the total number of tracked targets is large and the computer resource utilization rate is low, meeting the third preset condition, the system determines that it has the capability for parallel processing. At this time, the system allocates an independent computing thread for the tracking strategy of each target, enabling it to run in parallel, shortening the total time required to process all tracked targets, significantly improving the efficiency of multi-target concurrent processing, and thus ensuring stable, efficient, and flexible displacement detection in multi-target tracking scenarios.

[0058] When the total number of tracked targets is small or computer resource utilization is high, and the third preset condition is not met, the system switches to a single-threaded sequential execution mode. In this mode, all tracking strategies run sequentially according to preset execution priorities, effectively avoiding the additional system overhead caused by creating and managing multiple threads in low-load or resource-constrained scenarios, and preventing system lag or instability caused by resource contention. Therefore, this implementation allows the system to dynamically and intelligently switch between single-threaded and multi-threaded execution modes based on real-time operating conditions, ensuring both processing efficiency in complex scenarios and system stability and resource conservation in simple scenarios or when resources are scarce.

[0059] In a preferred embodiment, before thread scheduling and execution in step S105, for each tracking target determined in step S103 to employ a preset tracker strategy, the system initializes an independent tracker instance. Specifically, this initialization process creates a CSRT tracker instance by calling the cv2.TrackerCSRT_create() function in the opencv-python library, or creates a KCF tracker instance by calling the cv2.TrackerKCF_create() function, ensuring that each target has independent and non-interfering tracking resources.

[0060] Based on this, the system then makes a macro-mode judgment based on the real-time status of the total number of targets, CPU / GPU utilization, and frame rate: when there are a large number of targets and the system resources are sufficient, the system enters a multi-threaded mode, allocating an independent thread to each initialized tracker instance to run in parallel, thereby making full use of the multi-core processor architecture and significantly improving the overall efficiency of multi-target tracking.

[0061] When the number of targets is small or system resources are limited, a single-threaded mode is adopted, and all tracker instances are executed sequentially to avoid the additional overhead of thread creation, context switching and synchronization, and to ensure the stability of system operation.

[0062] In a preferred embodiment, a priority scheduling mechanism is also introduced. For critical targets that move at high speeds or are frequently occluded, the corresponding tracker instances will be given higher execution priority to ensure the tracking accuracy and real-time performance of these challenging targets.

[0063] When the system detects that a thread is running too slowly due to the complexity of the target, or that its resource consumption is abnormally high, it can automatically migrate the tracker instance that the thread is responsible for to other idle or low-load threads for execution. This effectively avoids dragging down the overall system performance due to a single point of bottleneck and ensures the smoothness and efficiency of the entire displacement monitoring process.

[0064] Further, step S105, which involves calculating the current coordinates of each tracking target in the current frame image using the current tracking strategy corresponding to each tracking target, includes: For any of the tracking targets, if the current tracking strategy for that tracking target is a preset tracker: The preset tracker is initialized based on the tracking target; The current frame image and the set of historical consecutive frame images are input to the preset tracker, and then the current coordinates of the tracked target in the current frame image are output through the preset tracker. For any of the tracking targets, if the current tracking strategy corresponding to the tracking target is a preset motion prediction model: The coordinates of the previous frame of the tracked target are obtained based on the historical continuous frame image set; According to the Kalman filter algorithm, a state transition matrix is ​​constructed based on the current displacement velocity, the displacement acceleration and the coordinates of the previous frame corresponding to the tracked target, and then the motion prediction model is constructed based on the state transition matrix; The current coordinates of the tracked target in the current frame image are obtained based on the motion prediction model.

[0065] In this implementation, for the strategy employing a preset tracker, the preset tracker, as an algorithm module, has the core function of accurately locating the target by analyzing its visual features within consecutive image frames. By inputting the current frame image and a set of historical consecutive frames into the initialized tracker, the system can dynamically match and lock onto the target using its appearance information. This method, which directly relies on image pixel information, enables the system to achieve accurate positioning even when the target's motion is complex, its appearance has changed partially, or there is strong background interference, thus ensuring the displacement monitoring accuracy for multi-target tracking in complex tracking scenarios.

[0066] On the other hand, for strategies employing pre-defined motion prediction models, this implementation avoids complex analysis by the tracker, thus saving computer resources. In this implementation, based on the Kalman filter algorithm, a state transition matrix is ​​constructed using only the target's previous frame coordinates, displacement velocity, and acceleration, quickly calculating the current coordinates. This avoids the computational overhead of tracker image processing, achieving rapid prediction of the target's position and significantly improving the system's target displacement monitoring response speed and operational efficiency in simple tracking scenarios.

[0067] Therefore, this implementation method adapts to the complex and ever-changing actual displacement monitoring application requirements by configuring differentiated execution mechanisms for different strategies, and ultimately achieves stable, efficient and flexible displacement detection.

[0068] In a preferred embodiment, in step S105, when a preset tracker strategy is adopted, the system inputs the current frame image to the corresponding tracker instance, obtains the new target position output by it, and calculates the center point of the rectangle as the precise coordinates of the target in the current frame.

[0069] When using a preset motion prediction model strategy, this preferred embodiment provides two specific prediction methods.

[0070] One type is the linear prediction model, which records the coordinates of the target center point in the two previous frames of the current frame image. and Based on this, the displacement increments Δx and Δy are calculated. Then, assuming the target moves at approximately a constant velocity, the displacement increments are extrapolated to the current frame to obtain the predicted position. .

[0071] This method extrapolates the displacement of the target position in the previous two frames of the current frame image based on the assumption that the target moves at an approximately uniform speed, thereby obtaining the predicted position region of the target.

[0072] Another more accurate Kalman filter prediction model constructs state variables containing the target's center position and velocity, and uses a state transition matrix to describe the inter-frame change relationship to estimate the target's predicted location region. Finally, the center point of this region is taken as the target's coordinates.

[0073] After calculating the current coordinates of all targets, step S106 is executed to compare the current coordinates of each target with its coordinates in the initial frame or a certain reference frame, thereby calculating its relative displacement and recording it.

[0074] In a preferred embodiment, a dynamic strategy adjustment and fault tolerance mechanism is introduced, which allows the system to evaluate the overall tracking status every few frames, including the tracker failure rate, CPU usage, and image blur level.

[0075] If a CSRT tracker matching a target is found to frequently lose targets, the tracker matching that target will be automatically downgraded to a KCF tracker.

[0076] If all trackers fail, the system seamlessly switches to a prediction mode that relies solely on the information from the previous frame. If a continuous increase in CPU load is detected, the system will proactively downgrade to single-threaded mode or shut down some target trackers, thereby ensuring that the entire displacement monitoring system can maintain high stability and reliability under complex and changing working conditions.

[0077] Please refer to Figure 2 A second aspect of the present invention provides an adaptive multi-mode switching image displacement monitoring system, comprising: The frame image processing module 100 is used to acquire the current frame image and a set of historical continuous frame images, and to identify several tracking targets in the set of historical continuous frame images; The target tracking analysis module 200 is used to obtain the current displacement velocity, displacement acceleration, trajectory smoothness and occlusion rate of each tracked target; The tracking strategy switching module 300 is used to determine the current tracking strategy for each tracking target based on the current displacement velocity, displacement acceleration, trajectory smoothness and occlusion rate of each tracking target. The current coordinate calculation module 400 is used to obtain the total number of tracking targets and the computer resource utilization rate. Based on the total number of tracking targets and the computer resource utilization rate, it performs thread scheduling for the current tracking strategy corresponding to each tracking target, and then calculates the current coordinates of each tracking target in the current frame image through the current tracking strategy corresponding to each tracking target. The displacement calculation module 500 is used to obtain the previous frame coordinates of each tracked target based on the historical continuous frame image set, and then calculate the relative displacement of each tracked target based on the current coordinates and the previous frame coordinates of each tracked target to complete image displacement monitoring.

[0078] Further, obtaining the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target includes: Based on the historical continuous frame image set, a first historical frame image and a second historical frame image are obtained, and then the inter-frame time difference between the first historical frame image and the second historical frame image is obtained; For any of the aforementioned tracking targets: Obtain the first coordinates of the tracked target in the first historical frame image, and obtain the second coordinates of the tracked target in the second historical frame image; The current displacement velocity is obtained based on the first coordinate, the second coordinate, and the inter-frame time difference; The historical displacement velocity of the tracked target is obtained based on the historical continuous frame image set, and then the displacement acceleration is obtained based on the current displacement velocity and the historical displacement velocity.

[0079] Further, obtaining the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target includes: For any of the aforementioned tracking targets: Based on the historical continuous frame image set, obtain several historical coordinates of the tracked target; Historical displacement trajectories are obtained based on several historical coordinates, the curvature of the historical displacement trajectories is calculated, and the smoothness of the trajectory is obtained based on the curvature. Based on the historical continuous frame image set, obtain several historical target images of the tracked target; Obtain the number of historical pixels corresponding to each historical target image, and then, based on the number of historical pixels corresponding to each historical target image and a preset occlusion ratio threshold, select a number of occluded target images from the number of historical target images. The occlusion rate is obtained by acquiring the proportion of all occluded target images in all historical target images.

[0080] Further, determining the current tracking strategy for each tracking target based on its current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate includes: For any of the aforementioned tracking targets: If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate satisfy the first preset condition, then a preset tracker is set as the current tracking strategy; If the occlusion rate meets the second preset condition, then the CSRT tracker is set as the preset tracker; otherwise, the KCF tracker is set as the preset tracker.

[0081] If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate do not meet the first preset condition, then a preset motion prediction model is set as the current tracking strategy.

[0082] The adaptive multi-mode switching image displacement monitoring method and system provided by the present invention have at least the following advantages compared with the prior art: This invention introduces an adaptive target tracking mode switching mechanism that intelligently switches between three modes—CSRT tracker, KCF tracker, and no tracker—based on factors such as the number of targets, motion characteristics, image stability, and system load in the actual application scenario. This mechanism overcomes the limitations of existing technologies that employ a single fixed tracking strategy, enabling the system to match the optimal processing scheme for different operating conditions. This maximizes computational efficiency and system adaptability while ensuring tracking accuracy.

[0083] This invention supports flexible selection of single-threaded or multi-threaded modes to execute tracking tasks based on the number of targets being tracked and the current system operating status. For single-target or light-load scenarios, single-threaded processing can be maintained to save resources; while when multiple targets need to be tracked simultaneously, a multi-threaded architecture can be automatically enabled, utilizing multi-core computing resources to execute multiple tracker tasks in parallel, thereby effectively avoiding problems such as frame rate drop, image stuttering, or tracking loss caused by serial processing, and significantly enhancing the scalability and stability of the system.

[0084] Because the system has the ability to analyze the target status of each frame of image in real time and can dynamically adjust the processing strategy, the present invention can still maintain high tracking accuracy and continuity in complex environments such as when the target is partially occluded, short-term loss occurs, or small displacement occurs. It effectively suppresses displacement jitter caused by target detection error and improves the accuracy and robustness of overall displacement measurement.

[0085] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0086] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; however, any combination of these technical features that does not contradict each other should be considered within the scope of this specification.

[0087] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the concept of this application, and these improvements and substitutions should also be considered within the scope of protection of this invention. Therefore, the scope of protection of this application should be determined by the appended claims.

Claims

1. An adaptive multi-mode switching image displacement monitoring method, characterized in that, include: Acquire the current frame image and a set of historical consecutive frame images, and identify several tracking targets in the set of historical consecutive frame images; Obtain the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate for each tracked target; Based on the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target, determine the current tracking strategy corresponding to each tracked target; Obtain the total number of tracked targets and computer resource utilization; Based on the total number of tracking targets and the computer resource utilization rate, thread scheduling is performed on the current tracking strategy corresponding to each tracking target, and then the current coordinates of each tracking target in the current frame image are calculated through the current tracking strategy corresponding to each tracking target. Based on the historical continuous frame image set, the coordinates of the previous frame of each tracked target are obtained. Then, based on the current coordinates and the coordinates of the previous frame of each tracked target, the relative displacement of each tracked target is calculated to complete the image displacement monitoring.

2. The adaptive multi-mode switching image displacement monitoring method according to claim 1, characterized in that, The acquisition of the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target includes: Based on the historical continuous frame image set, a first historical frame image and a second historical frame image are obtained, and then the inter-frame time difference between the first historical frame image and the second historical frame image is obtained; For any of the aforementioned tracking targets: Obtain the first coordinates of the tracked target in the first historical frame image, and obtain the second coordinates of the tracked target in the second historical frame image; The current displacement velocity is obtained based on the first coordinate, the second coordinate, and the inter-frame time difference; The historical displacement velocity of the tracked target is obtained based on the historical continuous frame image set, and then the displacement acceleration is obtained based on the current displacement velocity and the historical displacement velocity.

3. The image displacement monitoring method with adaptive multi-mode switching according to claim 2, characterized in that, The acquisition of the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target includes: For any of the aforementioned tracking targets: Based on the historical continuous frame image set, obtain several historical coordinates of the tracked target; Historical displacement trajectories are obtained based on several historical coordinates, the curvature of the historical displacement trajectories is calculated, and the smoothness of the trajectory is obtained based on the curvature. Based on the historical continuous frame image set, obtain several historical target images of the tracked target; The number of historical pixels corresponding to each historical target image is obtained, and then based on the number of historical pixels corresponding to each historical target image and a preset occlusion ratio threshold, a number of occluded target images are selected from a number of historical target images. The occlusion rate is obtained by acquiring the proportion of all occluded target images in all historical target images.

4. The image displacement monitoring method with adaptive multi-mode switching according to claim 1, characterized in that, The step of determining the current tracking strategy for each tracking target based on its current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate includes: For any of the aforementioned tracking targets: If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate satisfy the first preset condition, then a preset tracker is set as the current tracking strategy; If the occlusion rate meets the second preset condition, then the CSRT tracker is set as the preset tracker; otherwise, the KCF tracker is set as the preset tracker. If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate do not meet the first preset condition, then a preset motion prediction model is set as the current tracking strategy.

5. The image displacement monitoring method with adaptive multi-mode switching according to claim 1, characterized in that, The step of scheduling threads for the current tracking strategy corresponding to each tracking target based on the total number of tracking targets and the computer resource utilization rate includes: If the total number of tracked targets and the computer resource utilization rate meet the third preset condition, then an independent thread is allocated to each current tracking strategy so that all current tracking strategies can run in parallel; If the total number of tracked targets and the computer resource utilization rate do not meet the third preset condition, then all the current tracking strategies will run in a preset execution priority order.

6. The image displacement monitoring method with adaptive multi-mode switching according to claim 4, characterized in that, The step of calculating the current coordinates of each tracking target in the current frame image using the current tracking strategy corresponding to each tracking target includes: For any of the tracking targets, if the current tracking strategy for that tracking target is a preset tracker: The preset tracker is initialized based on the tracking target; The current frame image and the set of historical consecutive frame images are input to the preset tracker, and then the current coordinates of the tracked target in the current frame image are output through the preset tracker. For any of the tracking targets, if the current tracking strategy corresponding to the tracking target is a preset motion prediction model: The coordinates of the previous frame of the tracked target are obtained based on the historical continuous frame image set; According to the Kalman filter algorithm, a state transition matrix is ​​constructed based on the current displacement velocity, the displacement acceleration and the coordinates of the previous frame corresponding to the tracked target, and then the motion prediction model is constructed based on the state transition matrix; The current coordinates of the tracked target in the current frame image are obtained based on the motion prediction model.

7. An adaptive multi-mode switching image displacement monitoring system, characterized in that, include: The frame image processing module is used to acquire the current frame image and a set of historical continuous frame images, and to identify several tracking targets in the set of historical continuous frame images; The target tracking analysis module is used to obtain the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target; The tracking strategy switching module is used to determine the current tracking strategy for each tracking target based on the current displacement velocity, displacement acceleration, trajectory smoothness and occlusion rate of each tracking target. The current coordinate calculation module is used to obtain the total number of tracking targets and the computer resource utilization rate. Based on the total number of tracking targets and the computer resource utilization rate, it performs thread scheduling for the current tracking strategy corresponding to each tracking target, and then calculates the current coordinates of each tracking target in the current frame image through the current tracking strategy corresponding to each tracking target. The displacement calculation module is used to obtain the previous frame coordinates of each tracked target based on the historical continuous frame image set, and then calculate the relative displacement of each tracked target based on the current coordinates and the previous frame coordinates of each tracked target, thereby completing image displacement monitoring.

8. The adaptive multi-mode switching image displacement monitoring system according to claim 7, characterized in that, The acquisition of the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target includes: Based on the historical continuous frame image set, a first historical frame image and a second historical frame image are obtained, and then the inter-frame time difference between the first historical frame image and the second historical frame image is obtained; For any of the aforementioned tracking targets: Obtain the first coordinates of the tracked target in the first historical frame image, and obtain the second coordinates of the tracked target in the second historical frame image; The current displacement velocity is obtained based on the first coordinate, the second coordinate, and the inter-frame time difference; The historical displacement velocity of the tracked target is obtained based on the historical continuous frame image set, and then the displacement acceleration is obtained based on the current displacement velocity and the historical displacement velocity.

9. The adaptive multi-mode switching image displacement monitoring system according to claim 8, characterized in that, The acquisition of the current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate of each tracked target includes: For any of the aforementioned tracking targets: Based on the historical continuous frame image set, obtain several historical coordinates of the tracked target; Historical displacement trajectories are obtained based on several historical coordinates, the curvature of the historical displacement trajectories is calculated, and the smoothness of the trajectory is obtained based on the curvature. Based on the historical continuous frame image set, obtain several historical target images of the tracked target; The number of historical pixels corresponding to each historical target image is obtained, and then based on the number of historical pixels corresponding to each historical target image and a preset occlusion ratio threshold, a number of occluded target images are selected from a number of historical target images. The occlusion rate is obtained by acquiring the proportion of all occluded target images in all historical target images.

10. The adaptive multi-mode switching image displacement monitoring system according to claim 7, characterized in that, The step of determining the current tracking strategy for each tracking target based on its current displacement velocity, displacement acceleration, trajectory smoothness, and occlusion rate includes: For any of the aforementioned tracking targets: If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate satisfy the first preset condition, then a preset tracker is set as the current tracking strategy; If the occlusion rate meets the second preset condition, then the CSRT tracker is set as the preset tracker; otherwise, the KCF tracker is set as the preset tracker. If the current displacement velocity, the displacement acceleration, the trajectory smoothness, and the occlusion rate do not meet the first preset condition, then a preset motion prediction model is set as the current tracking strategy.

Images