基于改进Tracktor的行人多目标跟踪算法

在多目标视频跟踪中,针对受交互遮挡等影响导致检测偏差从而致使目标身份丢失的问题,提出一种基于改进Tracktor的行人多目标跟踪算法DUTracktor。在检测框回归中设计一个动态更新模块,利用孪生网络对建议框进一步检测定位;利用时序信息增强模块更新当前帧更适合的模板,建立全局上下文关系;并通过像素相关进行特征融合,从而增强目标边缘信息和尺度信息;利用相机运动补偿和融合相似矩阵构建二级关联跟踪机制,建立检测框和轨迹更强大的关联性,提高目标跟踪的鲁棒性。在公开的MOT16数据集上进行实验测试,并与当前主流算法相比,该算法跟踪精度表现较优,具有良好的鲁棒性,FPS稳定在24帧。

BEVTrack:基于难例挖掘训练的端到端三维多目标跟踪方法

多目标跟踪已经成为自动驾驶系统中的一个关键组成部分,其目的是在连续的视频流与点云流中识别、定位并标识所有感兴趣的目标。目前三维多目标跟踪方法多依赖人工多阶段调参以保证整体跟踪性能,难以对复杂遮挡或运动进行有效建模。而现有的三维端到端多目标跟踪方法,如MUTR等,精度普遍较低。其核心原因为三维空间中的特征聚合和感知相对于二维图像更具挑战性,简单的网络难以实现复杂的三维特征聚合,并大量的噪声信息与难例信息干扰严重,影响模型的特征提取能力。针对以上问题,本文提出了一种基于难例挖掘训练的端到端多目标跟踪框架BEVTrack。针对三维特征关联问题,本文设计了基于鸟瞰图(BEV)位置编码的三维跟踪查询。通过基于BEV特征的三维跟踪查询,本文方法能够更好地将跟踪查询与实际三维特征进行有效关联,从而大幅度提升了跟踪精度。同时,模型依靠BEV数据进行特征关联,仅需轻量化的网络便可以实现快速有效的跟踪。针对数据噪声问题,本文提出了面向多目标跟踪的难例挖掘训练,通过针对检测难例与跟踪难例分别处理,训练模型去除检测错误噪声与跟踪匹配的能力,从而提升在真实场景下模型处理噪声信息与难例干扰的能力。在实验结果方面,基于Nuscenes数据集,我们进行了大量的对比实验与模型消融实验,实验结果证明本文的方法在该数据集上取得了领先的性能。

基于改进CStrack关联策略的多目标跟踪算法

针对复杂场景下目标外观变化明显、运动不规律易导致轨迹中断和身份切换频繁等问题,从重识别(Re-Identification,Re-ID)特征、数据关联和插值等方面对跟踪器进行改进,提出基于改进CStrack关联策略的多目标跟踪算法。使用外观特征更新模块,减小因视角改变、目标移动导致特征剧烈变化而产生的影响,增强特征间的关联。提出二次关联方法,根据高低置信度检测结果的特点,使用不同的度量方式进行二次关联:第一次关联使用IoU距离融合外观特征作为关联的代价矩阵,第二次使用扩展IoU关联,缓解运动估计偏差、外观不可区分导致度量失效的问题;采用高斯回归算法,考虑运动信息,通过插值补偿漏检。在MOT17、MOT20数据集上进行测试,跟踪精度分别达到73.9%、64.2%。实验结果表明,该方法在跟踪精度上有明显优势,能够较好地适应复杂场景。

基于Bytetrack的多目标跟踪算法在斑马鱼毒性行为识别中的应用

利用计算机视觉技术识别斑马鱼(Danio rerio)在不同污染物暴露下的行为变化是水质毒性评价的常用方法之一,但传统方法存在效率低、面对遮挡和复杂环境时性能差等缺陷。针对这些问题,本研究使用基于Bytetrack的多目标跟踪算法追踪斑马鱼在4种污染物(Zn、Pb、Cr和苯酚)暴露2 h后的行为变化,对斑马鱼在4种浓度梯度中的平均速度、最大速度、最低速度、平均碰撞次数和行为轨迹等指标进行分析。结果显示,算法的追踪精度、漏检率和检测时间(每300帧)分别能达到90.26%、16.33%和0.19 min,检测时间和精度相比于传统目标检测方法有较大提升。同时,根据污染物不同,该方法能准确识别特定污染物环境中斑马鱼相应的运动状态及轨迹变化,可实现精确识别和实时响应,在鱼类毒性行为识别领域具有重要参考意义。

基于自编码器结构与改进Bytetrack的低光照行人检测及跟踪算法

针对夜间低光照场景下目标特征提取困难和跟踪不稳定的问题,提出了基于自编码器结构及改进Bytetrack的多目标行人检测及跟踪算法。在检测阶段,基于YOLOX(you only look once X)搭建多任务自编码变换模型框架,以一种自监督的方式考虑物理噪声模型和图像信号处理(image signal processing,ISP)的过程,通过对真实光照退化变换过程进行编码与解码学习内在视觉结构,并基于这种表示通过解码边界框坐标与类实现目标检测任务。为了抑制背景噪声的干扰,在目标解码器颈部网络引入自适应特征融合模块ASFF。跟踪阶段,基于Bytetrack算法进行改进,将基于Tranformer重识别网络提取到的外观嵌入信息与NSA卡尔曼滤波获得的运动信息通过自适应加权的方法完成数据关联,并通过Byte两次匹配的算法完成夜间行人的跟踪。在自建夜间行人检测数据集上测试检测模型的泛化能力,mAP@0.5达到了94.9%,结果表明本文的退化变换过程符合现实条件,具有良好的泛化能力。最后通过自建夜间行人跟踪数据集验证多目标跟踪性能,实验结果表明,本文提出的夜间低光照行人多目标跟踪算法MOTA(multiple object tracking accuracy)为89.55%,IDF1(identity F1 score)为88.34%,IDs(ID switches)为15。与基准方法Bytetrack相比,MOTA提高了10.72%,IDF1提高了6.19%,IDs减少了50%。结果表明,本文提出的基于自编码结构及改进Bytetrack的多目标跟踪算法可以有效解决在夜间低光照场景下行人跟踪困难的问题。

改进YOLOv7+Bytetrack的小目标检测与追踪

近年来,目标检测技术已经相当成熟,但小目标检测一直是目标检测领域的一大挑战。为了解决这一问题,设计一种名为MFF-YOLOv7的小目标检测算法,该算法旨在提高小目标检测的准确率。设计级联双向特征金字塔KBiFPN,以及联合提出的多级感受野特征聚合模块MFA,来聚合浅层特征并增强特征的信息表达能力。为了解决小目标漏检问题,设计了新的解耦头和新的注意力机制。新的解耦头对小目标的检测能力更强,新的注意力机制可以重点关注感兴趣的小目标区域。引入了一种新的损失函数ECIOU,旨在加快模型的收敛速度。为了验证模型的性能,分别在三个小目标数据集上进行了实验。实验结果表明,MFF-YOLOv7算法提高了检测精度。同时,使用多目标追踪Bytetrack算法在MOT17和VisDrone2019-MOT两个多目标追踪数据集上对新模型进行了验证,进一步证明了其有效性。此外,MFF-YOLOv7算法在动态视频追踪中表现出了良好的性能。

Obstacle Avoidance and Multitarget Tracking of a Super Redundant Modular Manipulator Based on Bezier Curve and Particle Swarm Optimization

A super redundant serpentine manipulator has slender structure and multiple degrees of freedom.It can travel through narrow spaces and move in complex spaces.This manipulator is composed of many modules that can form different lengths of robot arms for different application sites.The increase in degrees of freedom causes the inverse kinematics of redundant manipulator to be typical and immensely increases the calculation load in the joint space.This paper presents an integrated optimization method to solve the path planning for obstacle avoidance and discrete trajectory tracking of a super redundant manipulator.In this integrated optimization,path planning is established on a Bezier curve,and particle swarm optimization is adopted to adjust the control points of the Bezier curve with the kinematic constraints of manipulator.A feasible obstacle avoidance path is obtained along with a discrete trajectory tracking by using a follow-the-leader strategy.The relative distance between each two discrete path points is limited to reduce the fitting error of the connecting rigid links to the smooth curve.Simulation results show that this integrated optimization method can rapidly search for the appropriate trajectory to guide the manipulator in obtaining the target while achieving obstacle avoidance and meeting joint constraints.The proposed algorithm is suitable for 3D space obstacle avoidance and multitarget path tracking.

Application of Multi-Scale Tracking Radar Echoes Scheme in Quantitative Precipitation Nowcasting

A new radar echo tracking algorithm known as multi-scale tracking radar echoes by cross-correlation （MTREC） was developed in this study to analyze movements of radar echoes at different spatial scales. Movement of radar echoes, particularly associated with convective storms, exhibits different characteristics at various spatial scales as a result of complex interactions among meteorological systems leading to the formation of convective storms. For the null echo region, the usual correlation technique produces zero or a very small magnitude of motion vectors. To mitigate these constraints, MTREC uses the tracking radar echoes by correlation （TREC） technique with a large ＂box＂ to determine the systematic movement driven by steering wind, and MTREC applies the TREC technique with a small ＂box＂ to estimate small-scale internal motion vectors. Eventually, the MTREC vectors are obtained by synthesizing the systematic motion and the small-scale internal motion. Performance of the MTREC technique was compared with TREC technique using case studies： the Khanun typhoon on 11 September 2005 observed by Wenzhou radar and a squall-line system on 23 June 2011 detected by Beijing radar. The results demonstrate that more spatially smoothed and continuous vector fields can be generated by the MTREC technique, which leads to improvements in tracking the entire radar reflectivity pattern. The new multi-scMe tracking scheme was applied to study its impact on the performance of quantitative precipitation nowcasting. The location and intensity of heavy precipitation at a 1-h lead time was more consistent with quantitative precipitation estimates using radar and rain gauges.

Dynamic cluster member selection method for multi-target tracking in wireless sensor network

Multi-target tracking(MTT) is a research hotspot of wireless sensor networks at present.A self-organized dynamic cluster task allocation scheme is used to implement collaborative task allocation for MTT in WSN and a special cluster member(CM) node selection method is put forward in the scheme.An energy efficiency model was proposed under consideration of both energy consumption and remaining energy balance in the network.A tracking accuracy model based on area-sum principle was also presented through analyzing the localization accuracy of triangulation.Then,the two models mentioned above were combined to establish dynamic cluster member selection model for MTT where a comprehensive performance index function was designed to guide the CM node selection.This selection was fulfilled using genetic algorithm.Simulation results show that this method keeps both energy efficiency and tracking quality in optimal state,and also indicate the validity of genetic algorithm in implementing CM node selection.

Modified joint probabilistic data association with classification-aided for multitarget tracking

Joint probabilistic data association is an effective method for tracking multiple targets in clutter, but only the target kinematic information is used in measure-to-track association. If the kinematic likelihoods are similar for different closely spaced targets, there is ambiguity in using the kinematic information alone; the correct association probability will decrease in conventional joint probabilistic data association algorithm and track coalescence will occur easily. A modified algorithm of joint probabilistic data association with classification-aided is presented, which avoids track coalescence when tracking multiple neighboring targets. Firstly, an identification matrix is defined, which is used to simplify validation matrix to decrease computational complexity. Then, target class information is integrated into the data association process. Performance comparisons with and without the use of class information in JPDA are presented on multiple closely spaced maneuvering targets tracking problem. Simulation results quantify the benefits of classification-aided JPDA for improved multiple targets tracking, especially in the presence of association uncertainty in the kinematic measurement and target maneuvering. Simulation results indicate that the algorithm is valid.

Kernel density estimation and marginalized-particle based probability hypothesis density filter for multi-target tracking

In order to improve the performance of the probability hypothesis density(PHD) algorithm based particle filter(PF) in terms of number estimation and states extraction of multiple targets, a new probability hypothesis density filter algorithm based on marginalized particle and kernel density estimation is proposed, which utilizes the idea of marginalized particle filter to enhance the estimating performance of the PHD. The state variables are decomposed into linear and non-linear parts. The particle filter is adopted to predict and estimate the nonlinear states of multi-target after dimensionality reduction, while the Kalman filter is applied to estimate the linear parts under linear Gaussian condition. Embedding the information of the linear states into the estimated nonlinear states helps to reduce the estimating variance and improve the accuracy of target number estimation. The meanshift kernel density estimation, being of the inherent nature of searching peak value via an adaptive gradient ascent iteration, is introduced to cluster particles and extract target states, which is independent of the target number and can converge to the local peak position of the PHD distribution while avoiding the errors due to the inaccuracy in modeling and parameters estimation. Experiments show that the proposed algorithm can obtain higher tracking accuracy when using fewer sampling particles and is of lower computational complexity compared with the PF-PHD.

基于改进ByteTrack算法的红外地面多目标跟踪方法

红外目标智能检测跟踪技术研究一直是同领域中的热点问题,尤其是在精确制导、海面监视和天空预警等方面。针对红外地面多目标跟踪场景中,由地面杂波干扰、多目标遮挡干扰、平台晃动等复杂场景造成的跟踪精度降低等问题,提出了一种基于改进ByteTrack算法的红外地面多目标跟踪方法。首先引用一种自适应调制噪声尺度的卡尔曼滤波器,缓解低质量检测对vanilla卡尔曼滤波器的影响;其次引入增强相关系数最大化算法对帧间图像进行配准,来补偿平台晃动产生的影响;然后增加了基于长短期记忆网络的运动模型,减小了卡尔曼滤波在非线性运动状态中产生的预测误差;最后引入连接模型和高斯平滑算法这两种轻量级离线算法来完善跟踪结果。在红外地面多目标数据集上进行了实验,结果表明,与Sort和Deepsort算法相比,改进算法的MOTA值分别提升了8.3%和10.2%,IDF1值分别提升了6.5%和5.6%。与同类算法相比,改进算法表现出了更好的有效性,在红外目标智能检测跟踪场景中会有较大应用。

SMC-PHD based multi-target track-before-detect with nonstandard point observations model

Detection and tracking of multi-target with unknown and varying number is a challenging issue, especially under the condition of low signal-to-noise ratio(SNR). A modified multi-target track-before-detect(TBD) method was proposed to tackle this issue using a nonstandard point observation model. The method was developed from sequential Monte Carlo(SMC)-based probability hypothesis density(PHD) filter, and it was implemented by modifying the original calculation in update weights of the particles and by adopting an adaptive particle sampling strategy. To efficiently execute the SMC-PHD based TBD method, a fast implementation approach was also presented by partitioning the particles into multiple subsets according to their position coordinates in 2D resolution cells of the sensor. Simulation results show the effectiveness of the proposed method for time-varying multi-target tracking using raw observation data.

A new algorithm of bearings-only multi-target tracking of bistatic system

Much research mainly focuses on the batch processing method （e.g. maximum likelihood method） when bearings-only multiple targets tracking of bistatic sonar system is considered. In this paper, the idea of recursive processing method is presented and employed, and corresponding data association algorithms, i.e. a multi-objective ant-colony-based optimization algorithm and an easy fast assignment algorithm are developed to solve the measurements-to-measurements and measurements-to-tracks data association problems of bistatic sonar system, respectively. Monte-Carlo simulations are induced to evaluate the effectiveness of the proposed methods.

Synchronization of Radar Observations with Multi-Scale Storm Tracking

The 3-D radar reflectivity data has become increasingly important for use in data assimilation towards convective scale numerical weather prediction as well as next generation precipitation estimation. Typically, reflectivity data from multiple radars are objectively analyzed and mosaiced onto a regional 3-D Cartesian grid prior to being assimilated into the models. One multi-radar observations is the synchronization of all of the scientific issues associated with the mosaic of the observations. Since radar data is usually rapidly updated （-every 5-10 min）, it is common in current multi-radar mosaic techniques to combine multiple radar＇ observations within a time window by assunfing that the storms are steady within the window. The assumption holds well for slow evolving precipitation systems, but for fast evolving convective storms, this assumption may be violated and the mosaic of radar observations at different times may result in inaccurate storm structure depictions. This study investigates the impact of synchronization on storm structures in multiple radar data analyses using a multi-scale storm tracking algorithm.

Multi-Object Tracking with Micro Aerial Vehicle

A simple yet efficient tracking framework is proposed for real-time multi-object tracking with micro aerial vehicles(MAVs). It's basic missions for MAVs to detect specific targets and then track them automatically. In our method, candidate regions are generated using the salient detection in each frame and then classified by an eural network. A kernelized correlation filter(KCF) is employed to track each target until it disappears or the peak-sidelobe ratio is lower than a threshold. Besides, we define the birth and death of each tracker for the targets. The tracker is recycled if its target disappears and can be assigned to a new target. The algorithm is evaluated on the PAFISS and UAV123 datasets. The results show a good performance on both the tracking accuracy and speed.

Research and Application of Multi-Target Tracking Based on GM-PHD Filter

In recent years, multi-target tracking technology based on Gaussian Mixture- Probability Hypothesis Density (GM-PHD) filtering has become a hot field of information fusion research. This article outlines the generation and development of multi-target tracking methods based on GM-PHD filtering, and the principle and implementation method of GM-PHD filtering are explained, and the application status based on GM-PHD filtering is summarized, and the key issues of the development of GM-PHD filtering technology are analyzed.

Augmented input estimation in multiple maneuvering target tracking

This paper presents augmented input estimation(AIE)for multiple maneuvering target tracking.Multi-target tracking(MTT)is based on two main parts,data association and estimation.In data association(DA),the best observations are assigned to the considered tracks.In real conditions,the number of observations is more than targets and also locations of observations are often so scattered that the association between targets and observations cannot be done simply.In this case,for general MTT problems with unknown numbers of targets,we present a Markov chain Monte-Carlo DA(MCMCDA)algorithm that approximates the optimal Bayesian filter with low complexity in computations.After DA,estimation and tracking should be done.Since in general cases,many targets can have maneuvering motions,then AIE is proposed to cover both the non-maneuvering and maneuvering parts of motion and the maneuver detection procedure is eliminated.This model with an input estimation(IE)approach is a special augmentation in the state space model which considers both the state vector and the unknown input vector as a new augmented state vector.Some comparisons based on the Monte-Carlo simulations are also made to evaluate the performances of the proposed method and other older methods in MTT.

Tracking consensus for nonlinear heterogeneous multi-agent systems subject to unknown disturbances via sliding mode control

We investigate the tracking control for a class of nonlinear heterogeneous leader-follower multi-agent systems（MAS）with unknown external disturbances. Firstly, the neighbor-based distributed finite-time observers are proposed for the followers to estimate the position and velocity of the leader. Then, two novel distributed adaptive control laws are designed by means of linear sliding mode（LSM） as well as nonsingular terminal sliding mode（NTSM）, respectively. One can prove that the tracking consensus can be achieved asymptotically under LSM and the tracking error can converge to a quite small neighborhood of the origin in finite time by NTSM in spite of uncertainties and disturbances. Finally, a simulation example is given to verify the effectiveness of the obtained theoretical results.