A moving target multi-view cooperative tracking method of a UAV under low-illumination conditions
By using collaborative observation and reliability analysis from multiple UAVs, combined with Kalman filtering technology, the stability and accuracy issues of multi-view UAV tracking under low-light conditions were resolved, achieving efficient tracking of continuous target trajectories.
Patent Information
- Application Number
- CN202610324806.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-17
- Publication Date
- 2026-06-16
Smart Images
Figure CN122219489A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of collaborative tracking technology, specifically to a multi-view collaborative tracking method for moving targets of UAVs under low-light conditions. Background Technology
[0002] In the field of UAV target tracking, multi-view cooperative tracking technology has become an important means to solve target motion estimation in complex scenarios. Traditional methods typically rely on a single UAV viewpoint for target imaging and state estimation, which suffers from problems such as blind spots, high imaging noise in low-light environments, sensitivity to target occlusion, and trajectory discontinuities. To overcome these limitations, multiple UAVs have been gradually introduced in recent years to form multi-view observations, improving tracking stability through complementary perspectives. However, even with multi-view observations formed by UAV cooperation under low-light conditions, problems remain, including inconsistent image quality, significant illumination-related effects on some perspectives, and the inability to automatically correct single-view estimation errors. These issues prevent existing multi-view tracking methods from stably outputting continuous and accurate target trajectories in low-light scenarios.
[0003] In existing technologies, there are shortcomings in reliability assessment: existing technologies mainly rely on empirical indicators or single image quality evaluation methods to judge imaging reliability, such as evaluating imaging quality by average brightness or brightness contrast. However, these methods are difficult to comprehensively reflect the low-light imaging characteristics of the target area and fail to consider the contrast relationship between different viewpoints, resulting in low-quality observations having a negative impact on multi-view fusion and target state estimation. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention provides a multi-view cooperative tracking method for moving targets by unmanned aerial vehicles (UAVs) under low-light conditions, thereby solving the problems mentioned in the background section.
[0005] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a multi-view cooperative tracking method for moving targets by an unmanned aerial vehicle (UAV) under low-light conditions, comprising the following steps: S1. Multiple drones simultaneously image the target to obtain a multi-view target image sequence; S2. Perform credibility analysis on the multi-view target image sequence to obtain the view credibility set; S3. Estimate the target state based on the viewpoint credibility set to obtain the target state result; S4. Perform multi-view combination based on the target state results to obtain the joint target state; S5. Adjust the UAV observation position according to the joint target status to obtain the cooperative viewpoint distribution; S6. Perform continuous multi-view tracking based on the cooperative viewpoint distribution to obtain a stable target trajectory.
[0006] To further optimize this technical solution, the credibility analysis in step S2 includes: Based on the obtained multi-view target image sequence, the target area is located for target images obtained by different UAVs at the same time point, the imaging characteristics under low illumination conditions are analyzed, and the images are compared with target images at adjacent time points to evaluate the imaging stability of each view. The credibility is calculated by combining the imaging table characteristics and imaging stability to obtain the credibility set of each UAV view.
[0007] To further optimize this technical solution, the credibility calculation includes:
[0008] in: :time Time Credibility of a drone's perspective; :time Time Imaging quality metrics from a drone's perspective; Number of drones; The reliability of the UAV's perspective is obtained by normalizing the imaging quality indicators.
[0009] To further optimize this technical solution, the imaging quality indicators include:
[0010] in: :time Time Brightness discernibility of the target area from the perspective of a drone; :time Time Imaging stability score of the target area from the perspective of a drone; : Trade-off coefficient; The imaging quality index is obtained by weighting the brightness discernibility and imaging stability scores of the target area from the perspective of the UAV.
[0011] To further optimize this technical solution, the brightness discernibility includes:
[0012] in: :time Time The average brightness of the target area as a whole from the perspective of a drone; :time Time The overall brightness dispersion of the target area from the perspective of a drone; Minimal constant; The brightness discernibility is calculated based on the overall average brightness and brightness dispersion of the target area from the perspective of the drone.
[0013] To further optimize this technical solution, the imaging stability score includes:
[0014] in: :time Time The degree of overall imaging fluctuation of the target area from the perspective of a drone; An imaging stability score is obtained by mapping the overall imaging fluctuation of the target area from the perspective of the UAV.
[0015] To further optimize this technical solution, the viewpoint target state estimation in step S3 includes: Based on the obtained set of viewpoint confidence, the single-view target image sequence under each UAV viewpoint is bound to the confidence value at each corresponding moment. Based on the single-view target image sequence, the Kalman filter target tracking technology is used to recursively estimate the target motion state, and the target state result corresponding to each viewpoint is obtained.
[0016] To further optimize this technical solution, the multi-view combination in step S4 includes: Based on the obtained target state results, select the target state results of each perspective at the current moment, and perform collaborative fusion of multi-perspective target state results based on the corresponding perspective credibility to obtain the joint target state.
[0017] To further optimize this technical solution, the adjustment of the UAV observation position in step S5 includes: Based on the obtained joint target status, the observation space distribution of the UAVs is analyzed through the relative spatial relationship between each UAV and the target. The perspective distribution with observation redundancy and insufficient observation is identified, and the observation position of the UAVs is adjusted to form a cooperative perspective distribution.
[0018] To further optimize this technical solution, the continuous multi-view tracking in step S6 includes: Based on the obtained cooperative perspective distribution, the target is continuously tracked using multi-view observation results to obtain the joint target state at each time moment. The states are then correlated in chronological order to form a stable target trajectory.
[0019] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, they implement the steps of a multi-view cooperative tracking method for moving targets of a UAV under low illumination conditions as described in the first aspect of the present invention.
[0020] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of a multi-view cooperative tracking method for a moving target of an unmanned aerial vehicle under low illumination conditions as described in the first aspect of the present invention.
[0021] Compared with existing technologies, this invention provides a multi-view cooperative tracking method for moving targets of UAVs under low-light conditions, involving machine learning and deep learning technologies, and has the following beneficial effects: This method for multi-view cooperative tracking of moving targets by UAVs under low illumination conditions systematically calculates and normalizes the imaging credibility of each viewpoint through credibility analysis. This reduces the impact of single-view errors on multi-view fusion under low illumination conditions, improves the accuracy and reliability of multi-view fusion, and enhances tracking performance in low illumination scenarios. Attached Figure Description
[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0023] Figure 1 This is a flowchart illustrating a multi-view cooperative tracking method for moving targets by UAVs under low-light conditions proposed in this invention. Detailed Implementation
[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0026] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.
[0027] Example 1: Reference Figure 1 This is the first embodiment of the present invention, which provides a multi-view cooperative tracking method for moving targets of UAVs under low-light conditions, including the following steps: S1. Multiple drones simultaneously image the target to obtain a multi-view target image sequence.
[0028] In this embodiment, the synchronous imaging includes: In low-light environments, when a single UAV images a moving target, limitations imposed by sensor exposure capabilities and ambient lighting conditions often result in images with insufficient brightness, increased noise, blurred edges, and loss of local details. These issues directly lead to false detections, missed detections, or unstable state estimations during subsequent target detection and tracking. Furthermore, moving targets in real-world scenarios typically exhibit continuous motion. If observation relies solely on a single viewpoint, the single viewpoint will struggle to provide sustained, effective information should the target become occluded, rapidly change direction, or move away from the line of sight, causing tracking interruptions. Therefore, under low-light conditions, simply improving the performance of a single sensor or the quality of a single-view image is insufficient to fundamentally solve the tracking stability problem.
[0029] Under low illumination conditions, a multi-UAV collaborative observation system is constructed. Under the constraint of a unified time reference, each UAV performs continuous synchronous imaging of the same moving target from different spatial locations, obtaining multi-view target images in the same time series. These images are then arranged in chronological order to form a multi-view target image sequence, thus providing a data foundation for subsequent steps. This ensures that all subsequent processing steps are based on the same time reference, avoiding inconsistencies caused by time misalignment.
[0030] The simultaneous imaging steps include: Constructing a multi-UAV collaborative observation system: Under low-light conditions, a single UAV has difficulty obtaining a stable target image throughout the tracking process due to its limited observation angle and imaging information. Therefore, it is necessary to deploy at least two UAVs, each with independent flight capability and stable hovering capability to avoid additional imaging blur caused by flight jitter or unstable attitude. They are also equipped with visible light imaging sensors. All UAVs complete basic communication connections before the mission begins to ensure that subsequent time synchronization signals can be reliably distributed. Determine low-light operating conditions: When performing imaging tasks, the operating environment is limited to illumination conditions below the standard exposure limit of the imaging sensor. This ensures that the obtained images truly reflect the imaging characteristics under low-light conditions, rather than the results under ideal lighting conditions. This helps to verify the effectiveness of the multi-view collaborative mechanism in real low-light scenarios. Establish a unified time reference: Establish a unified time reference signal among multiple drones, and each drone performs image imaging operations according to this time reference to ensure that each drone images the target area at the same point in time, so that the target images from different perspectives are strictly aligned in the time dimension. Multi-view synchronous imaging: Under the constraint of a unified time reference, each UAV starts from its own spatial position and continuously images the same moving target. Since there are differences in spatial position, altitude and observation direction of the UAVs, the images obtained by different UAVs naturally form multiple perspectives, thus preserving the real observation differences of the target from different perspectives and providing a rich source of information for subsequent collaborative processing. Forming a multi-view target image sequence: Each UAV arranges the target images obtained by continuous imaging in chronological order to form its own target image sequence. Images from different UAVs with the same time mark are grouped together to form a multi-view target image sequence.
[0031] S2. Perform credibility analysis on the multi-view target image sequence to obtain the view credibility set.
[0032] In this embodiment, the credibility analysis includes: Under low-light conditions, even when multiple drones simultaneously image the same moving target, the quality of the target images obtained from different viewpoints still varies significantly due to differences in the spatial position, observation direction, target distance, and background illumination distribution of each drone. This difference is further amplified under low-light conditions, resulting in some viewpoints showing clear target outlines while others show noise-induced blurring, blurred edges, or partial indistinguishability. If, in subsequent processing, all images acquired from all viewpoints are considered equally reliable, the erroneous information contained in low-quality viewpoints will directly participate in target state estimation and multi-view fusion, thereby reducing the stability and accuracy of the overall tracking results.
[0033] Based on the obtained multi-view target image sequence, the target region is located for target images obtained by different UAVs at the same time point. The imaging characteristics under low-light conditions are analyzed and compared with target images at adjacent time points to evaluate the imaging stability of each view. The credibility is calculated by combining the imaging table characteristics and imaging stability to obtain the credibility set of each UAV view. This provides credibility constraints for multi-view collaboration, improves the accuracy and reliability of multi-view fusion, avoids the dominant influence of low-quality views on the overall tracking results, improves target tracking performance in low-light scenarios, and enhances the stability and accuracy of the overall tracking results.
[0034] Credibility analysis methods include: Extract target images from various perspectives: From the multi-view target image sequence obtained in step S1, extract target images obtained by different UAVs at the same time point to ensure that the analyzed images all correspond to the same motion state of the target; Analyze the imaging characteristics of the target area: For the target image of each UAV, locate the target area and analyze the brightness distribution, grayscale changes and noise performance of the area to reflect the identifiability of the target area under low light conditions. Assess imaging stability: Compare target images obtained from the same viewpoint at adjacent time points to determine the appearance changes of the target area during continuous imaging, thereby assessing the imaging stability of the target from that viewpoint in a low-light environment. If the imaging performance of a certain viewpoint fluctuates greatly in consecutive moments, it indicates that its reliability in a low-light environment is low. Comprehensive imaging quality index: It integrates the identifiability of the target area with the imaging stability results to form an evaluation result that can reflect the overall imaging performance of the viewpoint under the current low-light conditions, so that the evaluation result has both spatial and temporal significance. Forming viewpoint credibility: The above evaluation results are normalized and mapped to form the credibility of the corresponding viewpoint, so that each UAV has a clear credibility value at the current moment, which is used to represent the reliability of the viewpoint in reflecting the true state of the target at the current moment, and makes the credibility of different viewpoints comparable and can be directly used for subsequent processing, avoiding the inappropriate influence of low-quality viewpoints in the subsequent single-view state estimation and multi-view fusion process. Construct a set of view credibility: Collect the view credibility of all UAVs at the current time point to form a set of view credibility as the output result. Each multi-view target image sequence at each time point has its corresponding set of view credibility.
[0035] Furthermore, the credibility calculation includes:
[0036] in: :time Time The credibility of each drone's perspective ranges from 0 to 1. The larger the value, the higher the relative credibility of that drone's perspective on the target's true state at the current moment. The sum of the credibility of all drone perspectives at the same moment is 1. :time Time The imaging quality indicators from the perspective of a drone reflect the clarity and temporal stability of the target, and are the direct basis for forming the credibility of the perspective. : Number of drones, representing the total number of drones participating in multi-view collaborative tracking, determined based on the drone deployment results in step S1, with a minimum value of 2; The reliability of the UAV's perspective is obtained by normalizing the imaging quality indicators.
[0037] Furthermore, the imaging quality indicators include:
[0038] in: :time Time The brightness discernibility of a target area in the view of a drone is used to measure whether the target has sufficient distinguishability in a noisy background under the corresponding view. The higher the brightness and the smaller the dispersion, the larger the brightness discernibility value, indicating that the target area is clearer and easier to identify. :time Time The imaging stability score of the target area in the view of each UAV is used to measure the imaging stability under the corresponding view. The range is (0,1]. The smaller the continuous imaging brightness change, the more stable the imaging, and the closer the stability score is to 1. When the imaging fluctuation is obvious, the stability score decreases, reflecting that the view is unstable under low light conditions. : The tradeoff coefficient is used to adjust the relative influence of target sharpness and imaging stability in the comprehensive index. The larger the value, the more it emphasizes the current image sharpness, and the smaller the value, the more it emphasizes time stability. It can be set according to historical experimental experience, and the range is (0,1]. The imaging quality index is obtained by weighting the brightness discernibility and imaging stability scores of the target area from the perspective of the UAV.
[0039] Furthermore, the brightness discernibility includes:
[0040] in: :time Time The average brightness of the target area as a whole from the perspective of each drone represents the brightness at time [time]. No. The average grayscale value of all pixels in the target area from each viewpoint reflects the overall brightness of the target area. The larger the value, the higher the overall brightness of the target area. :time Time The overall brightness dispersion of the target area from the perspective of a drone indicates the degree of dispersion of the pixel gray value in the target area relative to the average brightness, that is, the strength of brightness change. It reflects the impact of noise, uneven brightness or imaging instability on the target area under low light conditions. The larger the value, the more obvious the brightness fluctuation of the target area. Minimal constant: A constant much smaller than the normal brightness variation, used to prevent division by zero when the brightness dispersion approaches zero, and to ensure the stability of the model values; The brightness discernibility is calculated based on the overall average brightness and brightness dispersion of the target area from the perspective of the drone.
[0041] Furthermore, the imaging stability score includes:
[0042] in: :time Time The degree of overall imaging fluctuation of the target area from the perspective of a drone indicates the range of change in the average brightness of the target area at adjacent time points. It is used to reflect the imaging stability of the same perspective at continuous time. The larger the value, the more obvious the imaging fluctuation of the perspective under low light conditions. An imaging stability score is obtained by mapping the overall imaging fluctuation of the target area from the perspective of the UAV.
[0043] Furthermore, the average brightness includes:
[0044] in: :time Target area grayscale value of each pixel; The number of pixels in the target area; The average brightness is calculated by summing and averaging the grayscale values of all pixels within the target area from the drone's perspective.
[0045] Furthermore, the brightness dispersion includes:
[0046] Based on the pixel grayscale values of the target area from the drone's perspective, the standard deviation of the grayscale values relative to the average brightness is calculated to obtain the brightness dispersion.
[0047] Furthermore, the degree of imaging fluctuation includes:
[0048] :time Time The average brightness of the target area as a whole from the perspective of a drone; The degree of image fluctuation is calculated by subtracting the average brightness of the target area from two consecutive frames from the same viewpoint and taking the absolute value.
[0049] S3. Estimate the target state based on the viewpoint credibility set to obtain the target state result.
[0050] In this embodiment, the viewpoint target state estimation includes: Under low-light conditions, the ultimate goal of multi-view cooperative tracking is to obtain a stable, continuous, and reliable target motion state. However, multi-view cooperation does not mean that the single-view processing stage can be skipped. Without target state estimation at the single-view level, multi-view cooperation will lose its most basic source of state input. At the same time, in low-light environments, the imaging quality under different UAV views varies significantly. Even though the imaging reliability has been quantified by view reliability in step S2, different views still need to independently estimate the target state to avoid noise amplification caused by directly fusing the original observations.
[0051] Based on the obtained set of viewpoint confidence, the single-view target image sequence under each UAV viewpoint is bound to the confidence value at each corresponding time. Based on the single-view target image sequence, Kalman filter target tracking technology is used to recursively estimate the target motion state, thereby obtaining the target state result corresponding to each viewpoint. This isolates the direct impact of low-light noise on the multi-view fusion stage, reflects the position change trend of the target in the time series, and provides a foundation for subsequent multi-view fusion.
[0052] The steps for estimating the state of a target from a specific viewpoint include: Determine the single-view processing object: Under low illumination conditions, the quality of target images under different UAV perspectives varies greatly. If multi-view information is directly introduced in the state estimation stage, it is easy to mix the noise in the poor-quality perspective with the effective information in the better perspective, which leads to instability in the state estimation process. Therefore, for the multi-view target image sequence formed in step S1, each UAV is taken as the processing object, and the single-view target image sequence obtained by the UAV in the continuous time is extracted to ensure that each perspective is processed independently. Introducing viewpoint credibility: From the viewpoint credibility set obtained in step S2, read the credibility value of each moment corresponding to the current UAV viewpoint, and bind it to the current viewpoint to identify the source of the reliability of the viewpoint state estimation result; Perform single-view target state estimation: Under each UAV view, based on the target image sequence of that view, the Kalman filter target tracking technology is used to recursively estimate the target motion state. This technology utilizes the continuity of the target's motion in the time series to predict and update the target position change, thereby suppressing the influence of random noise on the state estimation under low illumination noise interference conditions and obtaining smooth and continuous target state results. Forming target state results with independent perspectives: For each UAV, output the target state estimation result under its corresponding perspective. This result can be continuously updated over time and corresponds one-to-one with the credibility information of that perspective, thereby avoiding confusion of state results from different perspectives in terms of time or source, so that subsequent multi-perspective collaborative fusion can accurately identify which perspective each state result comes from and what its credibility level is. Summarize the target state results from multiple perspectives: Organize the target state results obtained from all UAV perspectives and the corresponding perspective credibility into a unified set of target states from multiple perspectives on the same time axis, which will be the final output of this step.
[0053] S4. Perform multi-perspective combination based on the target state results to obtain the joint target state.
[0054] In this embodiment, the multi-view combination includes: Under low illumination conditions, single-view imaging is significantly affected by low illumination noise, viewpoint occlusion, and changes in observation distance. The reliability of target state estimation varies significantly under different viewpoints. Any single viewpoint may experience state estimation drift or failure in a short period of time. If a certain viewpoint is directly selected as the final result, the overall robustness of the method will be severely limited. Furthermore, if the results of multiple views are simply averaged without distinction, the viewpoint with poor imaging quality will cause unreasonable interference to the final result.
[0055] Based on the obtained target state results, the target state results of each viewpoint at the current moment are selected. Based on the corresponding viewpoint credibility, the target state results of multiple views are fused collaboratively to obtain a joint target state. In this way, the influence of each viewpoint in the fusion process is explicitly constrained by the viewpoint credibility, and the independent target state results of multiple views are integrated into a single, continuous target state output. This improves the stability of the tracking method under low illumination, occlusion and short-term observation failure conditions, and provides reliable input for subsequent steps.
[0056] The steps of the multi-perspective joint approach include: Unify the time benchmark for target status from multiple perspectives: For the target status results from each perspective output in step S3, select target status results from different UAV perspectives that correspond to the same time point, based on the current processing time, to ensure that the status of each perspective is comparable in the time dimension. Read the credibility information corresponding to each perspective: For the target state result of each perspective, synchronously read the corresponding perspective credibility value to form a one-to-one correspondence between the target state result and the credibility, and provide a weight basis for subsequent fusion. Constructing a multi-perspective collaborative fusion relationship: At the same point in time, the target state results from different perspectives are regarded as multi-source descriptions of the same real target state. Based on their credibility, the degree of participation of each perspective state in the fusion process is determined. The higher the credibility, the more reliable the target state result under that perspective is, and the greater its influence in the formation of the joint target state. Perspectives with lower credibility only participate in the fusion as auxiliary information, thereby avoiding excessive interference of low-quality observations on the overall results. Weighted fusion is performed: based on the credibility of each perspective, the corresponding target state results are fused to form a single joint target state. This joint target state integrates multi-perspective information in terms of spatial position and motion trend, while maintaining temporal continuity. This allows for full utilization of multi-perspective information under low illumination conditions, while effectively suppressing the negative impact of perspectives with large noise or unstable imaging, making the joint target state more stable and continuous in terms of spatial position and motion trend. Output the joint target state result: The joint target state obtained by fusion at the current moment is used as the final output of this step, which can be directly called by subsequent steps.
[0057] Furthermore, the collaborative fusion includes:
[0058] in: :time The joint target state at that time represents the unified motion state of the target obtained after integrating information from multiple UAV perspectives under low illumination conditions. :time Time The target state result from a single UAV perspective represents an estimated description of the target's current motion state based on a single UAV perspective and a continuous image sequence from that perspective. It reflects the target's spatial position and its motion trend over time under that perspective and is obtained through step S3. By weighted and collaboratively fusing the target state results from all perspectives at the same moment, a joint target state is obtained.
[0059] S5. Adjust the UAV observation position according to the joint target status to obtain the cooperative viewpoint distribution.
[0060] In this embodiment, the adjustment of the UAV observation position includes: Under low-light conditions, the multi-view imaging quality of UAVs targeting moving targets is not only affected by sensor performance but also strongly depends on the relative spatial relationship between the UAV and the target, such as observation distance, observation orientation, and the degree of spatial complementarity between viewpoints. In step S4, a stable and continuous joint target state is obtained through multi-view collaborative fusion, which can accurately reflect the target's current position and movement trend. However, if the UAV's observation positions remain in their initial or static distribution, some UAV viewpoints may gradually deteriorate as the target moves, leading to problems such as viewpoint overlap, increased occlusion, or unreasonable imaging distances, thereby weakening the subsequent multi-view imaging quality.
[0061] Based on the obtained joint target state, the spatial distribution of UAV observations is analyzed through the relative spatial relationship between each UAV and the target. Viewpoint distributions with observation redundancy and insufficient observation are identified, and the observation positions of the UAVs are adjusted to form a cooperative viewpoint distribution. This transforms the target state estimation results into constraints on the spatial distribution of UAVs, improves the multi-view imaging conditions in subsequent moments, avoids long-term concentration of UAV viewpoints or degradation to single-direction observation, and significantly improves long-term tracking stability under low-light conditions.
[0062] Methods for adjusting the observation position of a drone include: Read the joint target state: Read the joint target state at the current moment from step S4. This state is used to characterize the target's current position in space and its direction of motion, serving as a unified reference for the UAV to adjust its observation position. Relative spatial relationship calculation: For each UAV, based on its current observation position and the joint target status, calculate the relative positional relationship between the UAV and the target, including relative distance and relative azimuth. The relative distance can be obtained by calculating the Euclidean distance using the standard three-dimensional distance formula, and the relative azimuth can be obtained by calculating the azimuth angle and pitch angle, which is used to determine whether the current observation conditions are reasonable. Identifying redundancy and deficiencies in multi-view distribution: Under low-light conditions, the advantages of multiple views depend on the complementarity between views. If multiple UAVs are concentrated in similar azimuths or at similar distances to observe the target, it will not only fail to significantly improve the overall imaging quality, but also waste view resources. If there is a long-term lack of observation in certain directions, it will lead to an increase in the uncertainty of the target state in these directions. By comprehensively considering the relative spatial relationship of each UAV, it can be determined whether there are multiple UAVs concentrated in similar observation azimuths, observation distances that are too close, or insufficient observation coverage in some directions, thereby identifying structural problems in the current view distribution. Adjusting UAV observation positions: Under the constraints of joint target status, the observation positions of some UAVs are adjusted. By changing the flight altitude, lateral position or heading of the UAVs, multiple UAVs can form a complementary observation relationship in space, avoiding overlapping viewpoints and maintaining continuous coverage of the target, thereby enhancing the robustness of overall imaging in low-light environments. Output cooperative view distribution: The adjusted UAV observation position is used as the output of this step to form a cooperative view distribution under the current target state constraints.
[0063] S6. Perform continuous multi-view tracking based on the cooperative viewpoint distribution to obtain a stable target trajectory.
[0064] In this embodiment, the continuous multi-view tracking includes: Under low-light conditions, outputting the target state only at a single moment without continuous tracking over time cannot meet the application requirements for moving target behavior analysis, trajectory prediction, or continuous monitoring. Furthermore, the effects of low-light environments are persistent and time-dependent; a single moment's joint target state cannot fully reflect the target's true trajectory. Without a continuous multi-moment collaborative tracking mechanism, errors may accumulate over long periods, leading to trajectory discontinuity or drift.
[0065] Based on the obtained cooperative viewpoint distribution, the target is continuously tracked using multi-view observation results to obtain the joint target state at each time point. The states are then correlated in chronological order to form a stable target trajectory. This suppresses the impact of short-term single-view failures on trajectory continuity, improves long-term tracking stability under low-light conditions, and fully realizes the technical closed loop from multi-view imaging to trajectory output, significantly enhancing the target tracking capability in complex low-light environments.
[0066] Methods for implementing continuous multi-view tracking include: Acquiring continuous multi-view observation results: Under the cooperative view distribution constraint formed in step S5, multiple UAVs continuously image the target, forming multi-view observation results that are updated over time, providing stable input for continuous tracking; Single-view state update: During continuous tracking, the target state needs to be updated over time. If we rely only on the joint target state of the previous moment and ignore the latest observation results of each view, the state estimate will gradually deviate from the true target position. Therefore, for each UAV view, the target state under that view is updated based on the target observation results obtained at the moment, so that the target state of each view can evolve continuously over time. Multi-view joint target state update: By integrating the target state update results from different perspectives and utilizing the complementary relationship between multiple perspectives, the target state of each perspective is updated collaboratively, so that the state information from different perspectives can be mutually corrected, the impact of abnormal fluctuations from a single perspective on the overall state can be reduced, and the stability of the target state in the time dimension can be improved to obtain the joint target state. By repeatedly executing the collaborative update process at continuous time points, the joint target state at each time point that evolves over time can be obtained to construct a stable target trajectory. Construct and maintain the target trajectory: Associate the joint target states obtained at each time step in chronological order to form a continuous target motion trajectory. Smooth out any sudden changes, discontinuities or abnormal deviations that may occur in the trajectory to keep the trajectory stable in space and time. Output stable target trajectory: The resulting continuous target trajectory is the final output of this step. This trajectory reflects the actual motion of the target under low light conditions and can be directly used for subsequent applications.
[0067] Example 2: In practical applications, this invention can be used by drones to perform multi-view collaborative tracking of moving vehicles or pedestrians under low-light conditions, with urban nighttime street traffic monitoring and rapid accident response as a typical scenario.
[0068] In this implementation, multiple drones simultaneously image the monitored area along a preset cruise route, acquiring a multi-view image sequence covering the target. The drones are equipped with high-sensitivity imaging sensors, enabling them to acquire target images in low-light or complex lighting conditions, while simultaneously recording the spatial coordinates of each drone. By performing brightness analysis and imaging quality assessment on the target images captured by each drone, a confidence set for each viewpoint is calculated, reflecting the reliability of target tracking from each perspective. Based on this, the target state from each viewpoint is independently estimated and then fused with the results from other viewpoints to form a joint estimate of the target's position and movement trend.
[0069] By combining the joint target status and dynamically adjusting the spatial distribution of UAVs based on their current position, speed, and observation conditions, multiple UAVs form complementary and redundantly controlled cooperative perspectives, improving the observability and imaging quality of targets throughout the monitoring area. During continuous flight, the UAVs continuously collect target information based on the optimized cooperative perspective distribution. Through multi-view status updates and fusion, continuous tracking of moving targets is achieved, generating stable and continuous target trajectories. These trajectories can be directly used for traffic flow analysis, rapid nighttime accident location, and UAV autonomous cruise navigation assistance, providing real-time and accurate decision-making support for urban nighttime traffic management.
[0070] In practical applications, this invention can effectively solve problems such as unstable single-view tracking, broken target trajectory, and mutual interference of observation viewpoints under low illumination conditions, ensuring the stability and accuracy of continuous multi-view tracking, and realizing efficient monitoring and precise positioning of moving targets by UAVs.
[0071] Example 3: This embodiment also provides a computer device applicable to a multi-view cooperative tracking method for moving targets of UAVs under low illumination conditions, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the multi-view cooperative tracking method for moving targets of UAVs under low illumination conditions as proposed in the above embodiment.
[0072] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a multi-view cooperative tracking method for moving targets of UAVs under low-light conditions as proposed in the above embodiment.
[0073] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.
[0074] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0075] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0076] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.
[0077] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
[0078] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A multi-view cooperative tracking method for moving targets using a UAV under low-light conditions, characterized in that, Includes the following steps: S1. Multiple drones simultaneously image the target to obtain a multi-view target image sequence; S2. Perform credibility analysis on the multi-view target image sequence to obtain the view credibility set; S3. Estimate the target state based on the viewpoint credibility set to obtain the target state result; S4. Perform multi-view combination based on the target state results to obtain the joint target state; S5. Adjust the UAV observation position according to the joint target status to obtain the cooperative viewpoint distribution; S6. Perform continuous multi-view tracking based on the cooperative viewpoint distribution to obtain a stable target trajectory.
2. The method for multi-view cooperative tracking of moving targets by UAVs under low illumination conditions according to claim 1, characterized in that, The credibility analysis in step S2 includes: Based on the obtained multi-view target image sequence, the target area is located for target images obtained by different UAVs at the same time point, the imaging characteristics under low illumination conditions are analyzed, and the images are compared with target images at adjacent time points to evaluate the imaging stability of each view. The credibility is calculated by combining the imaging table characteristics and imaging stability to obtain the credibility set of each UAV view.
3. The method for multi-view cooperative tracking of moving targets by UAVs under low illumination conditions according to claim 2, characterized in that, The credibility calculation includes: ； in: :time Time Credibility of a drone's perspective; :time Time Imaging quality metrics from a drone's perspective; Number of drones; The reliability of the UAV's perspective is obtained by normalizing the imaging quality indicators.
4. The method for multi-view cooperative tracking of moving targets by UAVs under low illumination conditions according to claim 3, characterized in that, The imaging quality indicators include: ； in: :time Time Brightness discernibility of the target area from the perspective of a drone; :time Time Imaging stability score of the target area from the perspective of a drone; : Trade-off coefficient; The imaging quality index is obtained by weighting the brightness discernibility and imaging stability scores of the target area from the perspective of the UAV.
5. A multi-view cooperative tracking method for moving targets of a UAV under low illumination conditions according to claim 4, characterized in that, The brightness discernibility includes: ； in: :time Time The average brightness of the target area as a whole from the perspective of a drone; :time Time The overall brightness dispersion of the target area from the perspective of a drone; Minimal constant; The brightness discernibility is calculated based on the overall average brightness and brightness dispersion of the target area from the perspective of the drone.
6. The method for multi-view cooperative tracking of moving targets by UAVs under low illumination conditions according to claim 4, characterized in that, The imaging stability score includes: ； in: :time Time The degree of overall imaging fluctuation of the target area from the perspective of a drone; An imaging stability score is obtained by mapping the overall imaging fluctuation of the target area from the perspective of the UAV.
7. A multi-view cooperative tracking method for moving targets of a UAV under low illumination conditions according to claim 1, characterized in that, The viewpoint target state estimation in step S3 includes: Based on the obtained set of viewpoint confidence, the single-view target image sequence under each UAV viewpoint is bound to the confidence value at each corresponding moment. Based on the single-view target image sequence, the Kalman filter target tracking technology is used to recursively estimate the target motion state, and the target state result corresponding to each viewpoint is obtained.
8. A multi-view cooperative tracking method for moving targets of a UAV under low illumination conditions according to claim 1, characterized in that, The multi-view combination in step S4 includes: Based on the obtained target state results, select the target state results of each perspective at the current moment, and perform collaborative fusion of multi-perspective target state results based on the corresponding perspective credibility to obtain the joint target state.
9. A multi-view cooperative tracking method for moving targets of a UAV under low illumination conditions according to claim 1, characterized in that, The adjustment of the UAV observation position in step S5 includes: Based on the obtained joint target status, the observation space distribution of the UAVs is analyzed through the relative spatial relationship between each UAV and the target. The perspective distribution with observation redundancy and insufficient observation is identified, and the observation position of the UAVs is adjusted to form a cooperative perspective distribution.
10. A multi-view cooperative tracking method for moving targets of a UAV under low illumination conditions according to claim 1, characterized in that, The continuous multi-view tracking in step S6 includes: Based on the obtained cooperative perspective distribution, the target is continuously tracked using multi-view observation results to obtain the joint target state at each time moment. The states are then correlated in chronological order to form a stable target trajectory.