A multi-target gimbal dynamic tracking method and system

Abstract

The present application relates to the technical field of intelligent monitoring, in particular to a multi-target pan-tilt dynamic tracking method and system. The system comprises: a perception module for collecting multi-source perception data of multiple targets in a scene; a priority evaluation module for performing priority quantitative calculation and dynamic sorting of each target based on task requirements, perception data and real-time state; a pan-tilt control module for determining whether to trigger pan-tilt switching according to the target priority sorting result through a preset switching decision algorithm, and controlling the pan-tilt to perform steering and focusing actions; and a data storage module for storing target priority data, position coordinates and switching logs. Before triggering switching, the three-dimensional coordinates of the original target are obtained and stored in the data storage module, and when the recovery condition is met, the three-dimensional coordinates of the original target are retrieved from the data storage module to restore tracking. This technical solution can improve the tracking efficiency and accuracy of the pan-tilt in a multi-target scene.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology, specifically to a method and system for dynamic tracking of multiple targets using a PTZ camera. Background Technology

[0002] In multi-target concurrent scenarios such as substation inspection and industrial workshop monitoring, PTZ tracking systems serve as a key technological means to ensure the safe operation of equipment and timely detection of anomalies. They bear the important task of identifying key targets and flexibly switching tracking objects according to the actual situation. However, existing PTZ tracking technologies face many technical problems that urgently need to be solved in practical applications.

[0003] Currently, PTZ tracking technology is mainly divided into two categories: single-target tracking and multi-target switching tracking. Due to its inherent characteristics, single-target tracking technology can only lock onto and continuously track a single target. When multiple targets appear simultaneously in the monitoring scene, the non-tracking targets will be lost directly. It cannot meet the actual needs of multi-target concurrent scenarios such as substations, and its practicality is extremely low.

[0004] While multi-target switching tracking technology has addressed some of the limitations of single-target tracking, it still has many shortcomings. Taking the comparative document CN118276608B as an example, this technology achieves gimbal adjustment through environmental data acquisition, target detection and classification, and causal relationship analysis. Its core idea is to set tracking priorities based on the speed and direction of dynamic targets and optimize adjustments by predicting target behavior. However, in practical applications, this technical solution has revealed the following problems: When prioritizing targets, current technologies often focus solely on their dynamic characteristics, such as speed and direction, neglecting crucial factors like task requirements and the strength of perceived signals. For instance, in substation inspection tasks, "fault-related targets" possess core attributes, and their task relevance should be a key dimension for priority assessment. Simultaneously, the degree of anomaly in fault signals (perceived signal strength) directly reflects the target's importance. However, existing technologies fail to consider these factors, leading to situations where critical fault targets like "high-temperature switchgear" may be prioritized by "non-faulty moving targets" due to their lower priority based solely on speed and direction. This increases the risk of missed detections and poses a threat to equipment safety.

[0005] In practical applications, two extreme situations can easily occur when the gimbal switches: on the one hand, switching is too frequent, and even slight priority fluctuations can trigger gimbal switching, causing frequent interruptions in the tracking process and making it impossible to track the target stably and continuously; on the other hand, switching is delayed, and even if the priority of the key target is significantly higher than that of the currently tracked target, the system still does not respond to the switching command, making it impossible to track the key target in a timely manner and making it difficult to adapt to complex scenarios with multiple targets dynamically changing.

[0006] Existing technologies often use preset, fixed weights for target priority, making it impossible to dynamically adjust them based on the urgency of different tasks. For example, the priority assessment of the same target should differ between routine inspection tasks and emergency troubleshooting tasks. In emergency scenarios, the weight of "faulty targets approaching the robot" should be increased promptly to highlight their importance, but existing technologies cannot achieve this dynamic adjustment, potentially leading to the inability to track critical targets in a timely manner during emergencies, thus increasing safety risks.

[0007] Existing technologies lack clear thresholds for determining "moving / stationary targets" and methods for calculating "approaching / moving away" distances, leading to ambiguity in target status assessment and affecting the accuracy of gimbal tracking. Furthermore, the absence of a mechanism for recording the original target's position during gimbal switching necessitates rescanning the entire scene when re-tracking the original target after switching, significantly reducing inspection efficiency and failing to meet the demands for high-efficiency tracking in practical applications. Summary of the Invention

[0008] The purpose of this invention is to propose a multi-target gimbal dynamic tracking method and system, which can improve the efficiency and accuracy of gimbal tracking in multi-target scenarios.

[0009] To achieve the above objectives, in a first aspect, the present invention proposes a multi-target gimbal dynamic tracking system, comprising: The perception module is used to collect multi-dimensional, multi-source perception data of multiple targets within the scene; The priority evaluation module is used to perform priority quantification and dynamic sorting of each objective based on task requirements, perceived data and real-time status. The gimbal control module is used to determine whether to trigger gimbal switching based on the target priority sorting result and a preset switching decision algorithm, and to control the gimbal to perform turning and focusing actions. The data storage module is used to store target priority data, location coordinates, and switching logs; Before triggering the switch, the three-dimensional coordinates of the original target are acquired and stored in the data storage module. When the recovery conditions are met, the three-dimensional coordinates of the original target are retrieved from the data storage module to resume tracking.

[0010] Beneficial effects of the basic solution: This technical solution achieves real-time status scoring through the in-loop homomorphism of task status and sensing data, and drives the control weight parameters of the gimbal through the scoring; it dynamically sorts multi-dimensional information such as task weight, sensing signal quality, and real-time target status, which solves the problem of one-sided evaluation caused by traditional gimbal tracking relying on a single condition judgment, and can significantly reduce the probability of missed detection and misjudgment of key targets in complex scenarios.

[0011] By using a preset switching decision algorithm to quantitatively determine the switching trigger conditions, and intelligently deciding whether to turn based on the priority ranking results, the gimbal can effectively avoid problems such as frequent jitter, blind switching or delayed response caused by target interference, ensuring a smooth and stable tracking process and improving the overall tracking reliability.

[0012] The evaluation rules and decision-making logic can be flexibly adjusted according to the needs of different scenarios and tasks. It can adaptively match the optimal tracking strategy in different task modes such as routine monitoring, emergency response, and key inspection. Its adaptability to complex and ever-changing scenarios is significantly better than that of traditional solutions with fixed strategies.

[0013] Before switching to a new target, the original target's 3D coordinates are automatically saved. When the recovery conditions are met, the historical coordinates can be directly called to quickly resume tracking without having to re-scan the entire area. This greatly shortens the time spent on re-tracking, reduces screen interruptions and target loss, and improves the efficiency of multi-target round-robin and backtracking.

[0014] The data storage module fully records priority data, target coordinates, and switching logs, enabling the tracking process to be traceable and verifiable. This facilitates subsequent fault analysis, strategy optimization, and system debugging, thereby improving the system's intelligent management level.

[0015] As a feasible preferred embodiment, the sensing module includes an infrared thermal imager, an acoustic fingerprint sensor, a high-definition camera, and a lidar. The infrared thermal imager is used to collect target temperature data, the voiceprint sensor is used to collect target voiceprint data, the high-definition camera is used to collect target image data, and the lidar is used to collect target position coordinate data.

[0016] As a feasible preferred embodiment, the priority evaluation module calculates the target priority using the following formula. :

[0017] in, As the ultimate priority of the goal, , and These are the weighting coefficients for task requirements, perceived data, and real-time status, respectively. , and The scores are quantitative ratings for task requirements, perceived data, and real-time status.

[0018] As a feasible and preferred option, the weight coefficients of the task requirement dimension are determined based on the statistical analysis of historical failure data. Different target types correspond to different basic weights, which are dynamically adjusted according to the task urgency level. The quantitative score of the perceived data dimension is based on the magnitude of the deviation of the target perceived signal from the normal value, and is calculated using a deviation-score mapping function. The higher the deviation, the higher the score. The evaluation metrics for the real-time state dimension include target mobility, relative position, and whether it has entered a critical area. The relative position is determined by the distance change over multiple consecutive frames to indicate whether the target is approaching, moving away, or relatively stationary.

[0019] As a feasible and preferred solution, the switching decision algorithm in the gimbal control module includes a greedy algorithm, the specific logic of which is as follows: Calculate the priority of all targets in the scene in real time and determine the highest priority. Priority of the current tracking target Calculate the percentage of the difference:

[0020] when When the preset threshold is reached, a switching command is triggered.

[0021] As a feasible and preferred solution, when the real-time distance between the target and the gimbal is less than a preset distance threshold, regardless of... If a preset threshold is reached, the highest priority switch will be triggered directly.

[0022] As a feasible preferred solution, the gimbal control module dynamically adjusts the turning speed according to the angle difference between the target and the current gimbal orientation; the larger the angle difference, the faster the turning speed.

[0023] As a feasible and preferred solution, the gimbal control module adopts a target size-focal length mapping model, which calculates the optimal focal length by combining the target pixel size h acquired by the camera with the range d measured by the lidar. The mapping formula is:

[0024] in, As the reference focal length, The actual height of the target This is the baseline distance.

[0025] Secondly, this invention proposes a multi-target gimbal dynamic tracking method, characterized in that it is applied to the aforementioned multi-target gimbal dynamic tracking system, comprising: The sensing module collects temperature, voiceprint, image and location data of multiple targets. The sensing module includes an infrared thermal imager, a voiceprint sensor, a high-definition camera and a lidar. Based on task requirements, perceived data, and real-time status, priority quantification and dynamic sorting are performed on each objective. Based on the priority sorting results, a greedy algorithm is used to determine whether to trigger gimbal switching, and to control the gimbal to complete the turning and focusing actions. Before triggering the switch, the three-dimensional coordinates of the original target are acquired and stored in the data storage module. When the recovery conditions are met, the three-dimensional coordinates of the original target are retrieved from the data storage module to resume tracking. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the architecture of a multi-target gimbal dynamic tracking system. Detailed Implementation

[0027] To make the technical solution and advantages of this application clearer, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. It is understood that the specific embodiments described herein are only some embodiments of the present invention, and are only used to explain this application, not to limit it. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered isolated; they can be combined with each other to achieve better technical effects. The same reference numerals appearing in the accompanying drawings of the following embodiments represent the same features or components, and can be applied to different embodiments.

[0028] Furthermore, unless otherwise defined, the technical or scientific terms used in this invention description shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains.

[0029] The present invention will now be described in further detail with reference to the accompanying drawings.

[0030] This disclosure provides a multi-target gimbal dynamic tracking system, referring to... Figure 1 It includes a perception module, a priority evaluation module, a gimbal control module, and a data storage module.

[0031] The sensing module includes an infrared thermal imager, a voiceprint sensor, a high-definition camera, and a lidar, which respectively collect target temperature, voiceprint, image, and location data.

[0032] Specifically, in this embodiment, the infrared thermal imager has a temperature detection range of -20℃ to 300℃ and an accuracy of ±0.3℃, accurately detecting the target's temperature information to determine if the equipment is overheating or experiencing other malfunctions. The acoustic signature sensor has a detection range of 30dB to 120dB and an accuracy of ±1dB, capable of collecting the target's acoustic signature signal and analyzing its characteristics to determine if the equipment is making abnormal noises or other malfunctions. The high-definition camera has a detection range of 1920×1080 resolution and a frame rate of 30fps, acquiring clear images of the target for target identification and status assessment. The lidar has a ranging range of 0.5m to 100m and an accuracy of ±2cm, accurately measuring the target's position coordinates and providing accurate position information for gimbal tracking.

[0033] The priority assessment module has a built-in three-dimensional assessment model for the quantitative calculation and dynamic ranking of target priorities. The priority calculation formula is as follows:

[0034] in, The final priority of the objective (0-100 points). , and The weighting coefficients are for task requirements, perceived data, and real-time status (the sum of the three is 1). , and The scores are quantitative ratings (0-100 points) for the dimensions of task requirements, perceived data, and real-time status.

[0035] The specific implementations for each dimension are as follows: Task requirement dimension ( , ): Based on statistical analysis of historical experimental failure data, the probability of accidents caused by different target types is calculated, and the higher the probability, the higher the task weight. In actual inspection and tracking scenarios, such as substation scenarios, statistical analysis of failure data from the past 5 years shows that the probability of accidents caused by "high-temperature switchgear" is 35%, "abnormal noise transformer" is 18%, and "moving personnel" is 2%. Based on this, the basic weights are set.

[0036] Substation inspection basic weights: High temperature target =0.4, abnormal noise target =0.25, moving personnel =0.1, no fault-related target =0.05.

[0037] Adjust the weight coefficient according to the urgency of the task, and trigger the dynamic adjustment rule through the "urgency level" in the task instruction. Routine inspection (urgency level 1): The weight remains at the basic value; Fault warning inspection (urgency level 2): The weight of the fault-related target is increased by 20% (e.g., the high-temperature target becomes 0.48); Emergency repair inspection (urgency level 3): The weight of the fault-related target is increased by 30%, and an additional 0.1 weight is superimposed on the "target approaching the robot".

[0038] According to the matching degree score of the target and the current task, a perfect match (such as the high-temperature switch cabinet in the equipment fault detection task) gets 100 points, a partial match (such as the abnormal sound transformer) gets 80 points, and no match gets 20 points.

[0039] Perceived data dimension ( , ): The evaluation indicators include signal strength, that is, the amplitude of the target perception signal deviating from the normal value, including the temperature deviation amplitude, the sound source decibel value, the brightness contrast, etc.

[0040] Basic weight = 0.3, which is联动提升 with the task requirement weight in the emergency scenario, and the maximum does not exceed 0.4.

[0041] Use the deviation - score mapping function to calculate the quantization score. The higher the deviation, the higher.

[0042] Temperature signal: Assume that the normal temperature range of the equipment is T±5℃, and the deviation amplitude is ; ΔT≤5℃ gets 20 points, 5℃ < ΔT≤10℃ gets 50 points, 10℃ < ΔT≤20℃ gets 80 points, ΔT>20℃ gets 100 points.

[0043] Voiceprint signal: The normal operating sound level of the equipment is below 60dB. The decibel value L≤60dB gets 20 points, 60dB < L≤80dB gets 50 points, 80dB < L≤100dB gets 80 points, L>100dB gets 100 points.

[0044] Real-time status dimension ( , ): The evaluation indicators include the target status, including mobility (moving / stationary), relative position (approaching / leaving), and size change (whether entering the critical area).

[0045] Basic weight = 0.3, which is increased to 0.35 at most in the emergency scenario.

[0046] Target status judgment logic: It should be noted that the term "联动提升" in the original text is not accurately translated as there is no exact equivalent in English. A more appropriate expression might be "linked increase" or "jointly increased", but without more context, the translation here tries to convey the general meaning. Also, the "<0000...>" tags are left unchanged as required. Movement / Stationary Determination: Target position coordinates are acquired continuously for 3 frames using a lidar system, and movement speed is calculated.

[0047] in, The interval between two frames, =0.033s, set the velocity threshold v=0.1m / s, v>v is judged as a moving target, v≤v is a stationary target; Approach / Distance Judgment: Calculate the real-time distance d between the target and the gimbal (robot), based on the distance change over 5 consecutive frames.

[0048] A value greater than 0.5m is considered an approximation. <-0.5m indicates a distance away; otherwise, it indicates relative stillness. The approach distance threshold is set to... =3m, d < d High priority is triggered when the value is 0; Critical area identification: The coordinate range of the critical equipment area in the substation is preset. If the center point of the target falls within this range, it is determined to be a critical area target.

[0049] The gimbal control module includes a gimbal drive unit and a switching decision unit. Based on the priority ranking result, it uses a greedy algorithm to determine whether to trigger gimbal switching and control the gimbal to complete turning and focusing actions.

[0050] Real-time calculation of the priority of all targets in the scene Find the highest priority Priority level of the current tracking target Calculate the percentage of the difference:

[0051] ≥20%, trigger the switching command; if multiple targets exist. satisfy ≥20%, then select The corresponding objective.

[0052] When the target distance d < d 0 (3m), regardless Whether the criteria are met will trigger a switch to the highest priority.

[0053] The gimbal is driven by a stepper motor, and its movement is based on the angle difference between the target and the current orientation of the gimbal. i Dynamically adjust steering speed. Angle difference. iWhen θ ≤ 30°, the steering speed v = 60 / s; when 30° < θ ≤ 90°, v = 120° / s; when θ > 90°, v = 180° / s, ensuring rapid steering while avoiding overshoot.

[0054] Using a target size-focal length mapping model, the optimal focal length is calculated by combining the target pixel size h acquired by the camera with the range d measured by the LiDAR. The mapping formula is:

[0055] in, As the reference focal length, The actual height of the target This is the reference distance. For example, if the actual target height is 1.2m, the reference focal length is 10m. =50mm, when measuring distance =5m, pixel size When =200px, the calculation is as follows =60mm, the gimbal will automatically adjust to this focal length to ensure the target is clearly focused.

[0056] Before switching, the three-dimensional coordinates (X, Y, Z) of the original target are obtained through LiDAR, and stored in the data storage module in combination with the current timestamp and target type; at the same time, the target tracking status is marked as paused. When the new target tracking is completed or the priority is reduced, the coordinates can be retrieved directly by resuming tracking command, and the gimbal jumps to that position to re-identify without full-scene scanning.

[0057] The data storage module stores target priority data, location coordinates, and switching logs, and supports quick retrieval of the original target's location. When the PTZ needs to resume tracking the original target, it can quickly obtain the original target's location information from the data storage module, improving tracking efficiency.

[0058] Taking substation fault early warning inspection (emergency level 2) as an example, the collaborative implementation process of each module is demonstrated. The complete process is as follows: Step S1: Task initialization and weight configuration. The system receives the "fault warning inspection command, emergency level 2 triggers dynamic weight adjustment - high temperature target task weight W=0.4×1.2=0.48, abnormal noise target W=0.25×1.2=0.3, sensing data weight W=0.3, real-time status weight W=0.3 (the sum of the three is corrected to 1.08, and normalized to 1); load the coordinates of key areas of the substation and normal parameters of the equipment (such as normal switchgear temperature T=35℃, normal transformer sound level ≤60dB).

[0059] Step S2: Multi-target perception and data acquisition. The perception modules work synchronously: the infrared thermal imager detects that the temperature of switch cabinet A is T=58℃, and the temperature of transformer B is normal; the acoustic sensor detects that the sound level of transformer B is L=85dB, and there is no abnormal acoustic signature in switch cabinet A; the lidar acquires the coordinates of three targets—switch cabinet A (X1,Y1,Z1), transformer B (X2,Y2,Z2), and mobile inspection personnel (X3,Y3,Z3); the camera simultaneously acquires images of each target.

[0060] Step S3, three-dimensional priority calculation, includes: Target 1 (Switchgear A, High Temperature): S = 100 points, S (Temperature deviation Δ) T =23℃) =100 points, S (stationary in the critical area) =80 points; P 1 = 0.48 × 100 + 0.3 × 100 + 0.22 × 80 = 48 + 30 + 17.6 = 95.6 points.

[0061] Target 2 (Transformer B, abnormal noise): S=80 points, S (sound level deviation of 25dB)=80 points, S (stillness)=30 points; P 2 = 0.3 × 80 + 0.3 × 80 + 0.4 × 30 = 24 + 24 + 12 = 60 points.

[0062] Target 3 (Moving Personnel): S=20 points, S(No Abnormal Signals)=20 points, S(Moving and Moving Away)=60 points; P 3 = 0.12 × 20 + 0.3 × 20 + 0.58 × 60 = 2.4 + 6 + 34.8 = 43.2 points.

[0063] Step S4, Greedy Algorithm Decision and Switching Control: Initially, the gimbal tracking target is a moving person. P (3 = 43.2 points), calculate the priority difference Δ P =( P 1- P 3) / P 3 = (95.6 - 43.2) / 43.2 ≈ 121% ≥ 20%, triggering the switching command.

[0064] PTZ control module: ① Records the coordinates (X3, Y3, Z3) of the moving personnel; ② Calculates the angle difference between switch cabinet A and the current orientation. i =45°, using a turning speed of 120° / s, the turning is completed in 0.375s; ③ Calculate the optimal focal length using a focal length adjustment algorithm. =55mm, focusing completed in 0.1s, total time 0.475s≤0.5s.

[0065] Step S5: Tracking and Status Update - The PTZ stabilizes and tracks switch cabinet A, updating its temperature and location data in real time; when the temperature of switch cabinet A drops to the normal range ( P (1 drops to 50 points), at which point transformer B has priority. P 2 = 60 points, Δ P =(60-50) / 50=20%, triggering the switch. The system retrieves the coordinates of transformer B to complete the rotation, and simultaneously records the position of switchgear A to achieve dynamic cyclic tracking.

[0066] The above content is merely an embodiment of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can improve and implement this solution based on the guidance provided in this application and their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A multi-target gimbal dynamic tracking system, characterized in that, include: The perception module is used to collect multi-source perception data of multiple targets within the scene; The priority evaluation module is used to perform priority quantification and dynamic sorting of each objective based on task requirements, perceived data and real-time status. The gimbal control module is used to determine whether to trigger gimbal switching based on the target priority sorting result and a preset switching decision algorithm, and to control the gimbal to perform turning and focusing actions. The data storage module is used to store target priority data, location coordinates, and switching logs; Before triggering the switch, the three-dimensional coordinates of the original target are acquired and stored in the data storage module. When the recovery conditions are met, the three-dimensional coordinates of the original target are retrieved from the data storage module to resume tracking.

2. The multi-target gimbal dynamic tracking system according to claim 1, characterized in that, The sensing module includes an infrared thermal imager, an acoustic sensor, a high-definition camera, and a lidar. The infrared thermal imager is used to collect target temperature data, the voiceprint sensor is used to collect target voiceprint data, the high-definition camera is used to collect target image data, and the lidar is used to collect target position coordinate data.

3. The multi-target gimbal dynamic tracking system according to claim 1, characterized in that, The priority evaluation module calculates the target priority using the following formula. : in, As the ultimate priority of the goal, , and These are the weighting coefficients for task requirements, perceived data, and real-time status, respectively. , and The scores are quantitative ratings for task requirements, perceived data, and real-time status.

4. The multi-target gimbal dynamic tracking system according to claim 3, characterized in that, The weighting coefficients for the task requirement dimension are determined based on statistical analysis of historical failure data. Different target types correspond to different basic weights, which are dynamically adjusted according to the task urgency level. The quantitative score of the perceived data dimension is based on the magnitude of the deviation of the target perceived signal from the normal value, and is calculated using a deviation-score mapping function. The higher the deviation, the higher the score. The evaluation metrics for the real-time state dimension include target mobility, relative position, and whether it has entered a critical area. The relative position is determined by the distance change over multiple consecutive frames to indicate whether the target is approaching, moving away, or relatively stationary.

5. A multi-target gimbal dynamic tracking system according to claim 1, characterized in that, The switching decision algorithm in the gimbal control module includes a greedy algorithm, the specific logic of which is as follows: Calculate the priority of all targets in the scene in real time and determine the highest priority. Priority of the current tracking target Calculate the percentage of the difference: when When the preset threshold is reached, a switching command is triggered.

6. A multi-target gimbal dynamic tracking system according to claim 5, characterized in that, When the real-time distance between the target and the gimbal is less than a preset distance threshold, regardless of If a preset threshold is reached, the highest priority switch will be triggered directly.

7. A multi-target gimbal dynamic tracking system according to claim 1, characterized in that, The gimbal control module dynamically adjusts the turning speed based on the angle difference between the target and the current gimbal orientation; the greater the angle difference, the faster the turning speed.

8. A multi-target gimbal dynamic tracking system according to claim 1, characterized in that, The gimbal control module employs a target size-focal length mapping model, calculating the optimal focal length by combining the target pixel size h acquired by the camera with the range measurement d from the LiDAR. The mapping formula is: in, As the reference focal length, For the actual target height, This is the baseline distance.

9. A multi-target gimbal dynamic tracking method, characterized in that, The system is applied to a multi-target gimbal dynamic tracking system as described in claims 1-9, comprising: The sensing module collects temperature, voiceprint, image and location data of multiple targets. The sensing module includes an infrared thermal imager, a voiceprint sensor, a high-definition camera and a lidar. Based on task requirements, perceived data, and real-time status, priority quantification and dynamic sorting are performed on each objective. Based on the priority sorting results, a greedy algorithm is used to determine whether to trigger gimbal switching, and to control the gimbal to complete the turning and focusing actions. Before triggering the switch, the three-dimensional coordinates of the original target are acquired and stored in the data storage module. When the recovery conditions are met, the three-dimensional coordinates of the original target are retrieved from the data storage module to resume tracking.

Images