End-to-End Joint Multi-Object Detection and Tracking for Intelligent Transportation Systems

Environment perception is one of the most critical technology of intelligent transportation systems(ITS).Motion interaction between multiple vehicles in ITS makes it important to perform multi-object tracking(MOT).However,most existing MOT algorithms follow the tracking-by-detection framework,which separates detection and tracking into two independent segments and limit the global efciency.Recently,a few algorithms have combined feature extraction into one network;however,the tracking portion continues to rely on data association,and requires com‑plex post-processing for life cycle management.Those methods do not combine detection and tracking efciently.This paper presents a novel network to realize joint multi-object detection and tracking in an end-to-end manner for ITS,named as global correlation network(GCNet).Unlike most object detection methods,GCNet introduces a global correlation layer for regression of absolute size and coordinates of bounding boxes,instead of ofsetting predictions.The pipeline of detection and tracking in GCNet is conceptually simple,and does not require compli‑cated tracking strategies such as non-maximum suppression and data association.GCNet was evaluated on a multivehicle tracking dataset,UA-DETRAC,demonstrating promising performance compared to state-of-the-art detectors and trackers.

Simultaneous Multi-vehicle Detection and Tracking Framework with Pavement Constraints Based on Machine Learning and Particle Filter Algorithm

Due to the large variations of environment with ever-changing background and vehicles with different shapes, colors and appearances, to implement a real-time on-board vehicle recognition system with high adaptability, efficiency and robustness in complicated environments, remains challenging. This paper introduces a simultaneous detection and tracking framework for robust on-board vehicle recognition based on monocular vision technology. The framework utilizes a novel layered machine learning and particle filter to build a multi-vehicle detection and tracking system. In the vehicle detection stage, a layered machine learning method is presented, which combines coarse-search and fine-search to obtain the target using the AdaBoost-based training algorithm. The pavement segmentation method based on characteristic similarity is proposed to estimate the most likely pavement area. Efficiency and accuracy are enhanced by restricting vehicle detection within the downsized area of pavement. In vehicle tracking stage, a multi-objective tracking algorithm based on target state management and particle filter is proposed. The proposed system is evaluated by roadway video captured in a variety of traffics, illumination, and weather conditions. The evaluating results show that, under conditions of proper illumination and clear vehicle appearance, the proposed system achieves 91.2% detection rate and 2.6% false detection rate. Experiments compared to typical algorithms show that, the presented algorithm reduces the false detection rate nearly by half at the cost of decreasing 2.7%–8.6% detection rate. This paper proposes a multi-vehicle detection and tracking system, which is promising for implementation in an on-board vehicle recognition system with high precision, strong robustness and low computational cost.

Surrounding Objects Detection and Tracking for Autonomous Driving Using LiDAR and Radar Fusion

Radar and LiDAR are two environmental sensors commonly used in autonomous vehicles,Lidars are accurate in determining objects’positions but significantly less accurate as Radars on measuring their velocities.However,Radars relative to Lidars are more accurate on measuring objects velocities but less accurate on determining their positions as they have a lower spatial resolution.In order to compensate for the low detection accuracy,incomplete target attributes and poor environmental adaptability of single sensors such as Radar and LiDAR,in this paper,an effective method for high-precision detection and tracking of surrounding targets of autonomous vehicles.By employing the Unscented Kalman Filter,Radar and LiDAR information is effectively fused to achieve high-precision detection of the position and speed information of targets around the autonomous vehicle.Finally,the real vehicle test under various driving environment scenarios is carried out.The experimental results show that the proposed sensor fusion method can effectively detect and track the vehicle peripheral targets with high accuracy.Compared with a single sensor,it has obvious advantages and can improve the intelligence level of autonomous cars.

Architectural Building Detection and Tracking in Video Sequences Taken by Unmanned Aircraft System （UAS）

An Unmanned Aircraft System （UAS） is an aircraft or ground station that can be either remote controlled manually or is capable of flying autonomously under the guidance of pre-programmed Global Positioning System （GPS） waypoint flight plans or more complex onboard intelligent systems. The UAS aircrafts have recently found extensive applications in military reconnaissance and surveillance, homeland security, precision agriculture, fire monitoring and analysis, and other different kinds of aids needed in disasters. Through surveillance videos captured by a UAS digital imaging payload over the interest areas, the corresponding UAS missions can be conducted. In this paper, the authors present an effective method to detect and extract architectural buildings under rural environment from UAS video sequences. The SIFT points are chosen as image features. The planar homography is adopted as the motion model between different image frames. The proposed algorithm is tested on real UAS video data.

Road boundary estimation to improve vehicle detection and tracking in UAV video

Video processing is one challenge in collecting vehicle trajectories from unmanned aerial vehicle(UAV) and road boundary estimation is one way to improve the video processing algorithms. However, current methods do not work well for low volume road, which is not well-marked and with noises such as vehicle tracks. A fusion-based method termed Dempster-Shafer-based road detection(DSRD) is proposed to address this issue. This method detects road boundary by combining multiple information sources using Dempster-Shafer theory(DST). In order to test the performance of the proposed method, two field experiments were conducted, one of which was on a highway partially covered by snow and another was on a dense traffic highway. The results show that DSRD is robust and accurate, whose detection rates are 100% and 99.8% compared with manual detection results. Then, DSRD is adopted to improve UAV video processing algorithm, and the vehicle detection and tracking rate are improved by 2.7% and 5.5%,respectively. Also, the computation time has decreased by 5% and 8.3% for two experiments, respectively.

An array of two periodic leaky-wave antennas with sum and difference beam scanning for application in target detection and tracking

An array of two substrate-integrated waveguide(SIW) periodic leaky-wave antennas(LWAs) with sum and difference beam scanning is proposed for application in target detection and tracking. The array is composed of two periodic LWAs with different periods, in which each LWA generates a narrow beam through the n=-1 space harmonic. Due to the two different periods for the two LWAs, two beams with two different directions can be realized, which can be combined into a sum beam when the array is fed in phase or into a difference beam when the array is fed 180?out of phase. The array integrated with 180?hybrid is designed, fabricated, and measured.Measurement results show that the sum beam can reach a gain up to 15.9 dBi and scan from-33.4?to 20.8?. In the scanning range, the direction of the null in the difference beam is consistent with the direction of the sum beam,with the lowest null depth of-40.8 dB. With the excellent performance, the antenna provides an alternative solution with low complexity and low cost for target detection and tracking.

Novel method for real-time detection and tracking of pig body and its different parts

Detection and tracking of all major parts of pig body could be more productive to help to analyze pig behavior.To achieve this goal,a real-time algorithm based on You Only Look At CoefficienTs(YOLACT)was proposed.A pig body was divided into ten parts:one head,one trunk,four thighs and four shanks.And the key points of each part were calculated by the novel algorithm,which was based mainly on combination of the Zhang-Suen thinning algorithm and Gravity algorithm.The experiment results showed that these parts of pig body could be detected and tracked,and their contributions to overall pig activity could also be sought out.The detect accuracy of the algorithm in the data set could reach up to 90%,and the processing speed to 30.5 fps.Furthermore,the algorithm was robust and adaptive.

Design of a wide-field target detection and tracking system using the segmented planar imaging detector for electro-optical reconnaissance

Detecting and tracking multiple targets simultaneously for space-based surveillance requires multiple cameras,which leads to a large system volume and weight. To address this problem, we propose a wide-field detection and tracking system using the segmented planar imaging detector for electro-optical reconnaissance. This study realizes two operating modes by changing the working paired lenslets and corresponding waveguide arrays： a detection mode and a tracking mode. A model system was simulated and evaluated using the peak signal-to-noise ratio method. The simulation results indicate that the detection and tracking system can realize wide-field detection and narrow-field, multi-target, high-resolution tracking without moving parts.

Semantic Segmentation and YOLO Detector over Aerial Vehicle Images

Intelligent vehicle tracking and detection are crucial tasks in the realm of highway management.However,vehicles come in a range of sizes,which is challenging to detect,affecting the traffic monitoring system’s overall accuracy.Deep learning is considered to be an efficient method for object detection in vision-based systems.In this paper,we proposed a vision-based vehicle detection and tracking system based on a You Look Only Once version 5(YOLOv5)detector combined with a segmentation technique.The model consists of six steps.In the first step,all the extracted traffic sequence images are subjected to pre-processing to remove noise and enhance the contrast level of the images.These pre-processed images are segmented by labelling each pixel to extract the uniform regions to aid the detection phase.A single-stage detector YOLOv5 is used to detect and locate vehicles in images.Each detection was exposed to Speeded Up Robust Feature(SURF)feature extraction to track multiple vehicles.Based on this,a unique number is assigned to each vehicle to easily locate them in the succeeding image frames by extracting them using the feature-matching technique.Further,we implemented a Kalman filter to track multiple vehicles.In the end,the vehicle path is estimated by using the centroid points of the rectangular bounding box predicted by the tracking algorithm.The experimental results and comparison reveal that our proposed vehicle detection and tracking system outperformed other state-of-the-art systems.The proposed implemented system provided 94.1%detection precision for Roundabout and 96.1%detection precision for Vehicle Aerial Imaging from Drone(VAID)datasets,respectively.

CGTracker:Center Graph Network for One-Stage Multi-Pedestrian-Object Detection and Tracking

Most current online multi-object tracking(MOT)methods include two steps:object detection and data association,where the data association step relies on both object feature extraction and affinity computation.This often leads to additional computation cost,and degrades the efficiency of MOT methods.In this paper,we combine the object detection and data association module in a unified framework,while getting rid of the extra feature extraction process,to achieve a better speed-accuracy trade-off for MOT.Considering that a pedestrian is the most common object category in real-world scenes and has particularity characteristics in objects relationship and motion pattern,we present a novel yet efficient one-stage pedestrian detection and tracking method,named CGTracker.In particular,CGTracker detects the pedestrian target as the center point of the object,and directly extracts the object features from the feature representation of the object center point,which is used to predict the axis-aligned bounding box.Meanwhile,the detected pedestrians are constructed as an object graph to facilitate the multi-object association process,where the semantic features,displacement information and relative position relationship of the targets between two adjacent frames are used to perform the reliable online tracking.CGTracker achieves the multiple object tracking accuracy(MOTA)of 69.3%and 65.3%at 9 FPS on MOT17 and MOT20,respectively.Extensive experimental results under widely-used evaluation metrics demonstrate that our method is one of the best techniques on the leader board for the MOT17 and MOT20 challenges at the time of submission of this work.

Towards Collaborative Robotics in Top View Surveillance:A Framework for Multiple Object Tracking by Detection Using Deep Learning

Collaborative Robotics is one of the high-interest research topics in the area of academia and industry.It has been progressively utilized in numerous applications,particularly in intelligent surveillance systems.It allows the deployment of smart cameras or optical sensors with computer vision techniques,which may serve in several object detection and tracking tasks.These tasks have been considered challenging and high-level perceptual problems,frequently dominated by relative information about the environment,where main concerns such as occlusion,illumination,background,object deformation,and object class variations are commonplace.In order to show the importance of top view surveillance,a collaborative robotics framework has been presented.It can assist in the detection and tracking of multiple objects in top view surveillance.The framework consists of a smart robotic camera embedded with the visual processing unit.The existing pre-trained deep learning models named SSD and YOLO has been adopted for object detection and localization.The detection models are further combined with different tracking algorithms,including GOTURN,MEDIANFLOW,TLD,KCF,MIL,and BOOSTING.These algorithms,along with detection models,help to track and predict the trajectories of detected objects.The pre-trained models are employed;therefore,the generalization performance is also investigated through testing the models on various sequences of top view data set.The detection models achieved maximum True Detection Rate 93%to 90%with a maximum 0.6%False Detection Rate.The tracking results of different algorithms are nearly identical,with tracking accuracy ranging from 90%to 94%.Furthermore,a discussion has been carried out on output results along with future guidelines.

Passive target tracking with intermittent measurement based on random finite set

In the tracking problem for the maritime radiation source by a passive sensor,there are three main difficulties,i.e.,the poor observability of the radiation source,the detection uncertainty(false and missed detections)and the uncertainty of the target appearing/disappearing in the field of view.These difficulties can make the establishment or maintenance of the radiation source target track invalid.By incorporating the elevation information of the passive sensor into the automatic bearings-only tracking(BOT)and consolidating these uncertainties under the framework of random finite set(RFS),a novel approach for tracking maritime radiation source target with intermittent measurement was proposed.Under the RFS framework,the target state was represented as a set that can take on either an empty set or a singleton; meanwhile,the measurement uncertainty was modeled as a Bernoulli random finite set.Moreover,the elevation information of the sensor platform was introduced to ensure observability of passive measurements and obtain the unique target localization.Simulation experiments verify the validity of the proposed approach for tracking maritime radiation source and demonstrate the superiority of the proposed approach in comparison with the traditional integrated probabilistic data association(IPDA)method.The tracking performance under different conditions,particularly involving different existence probabilities and different appearance durations of the target,indicates that the method to solve our problem is robust and effective.

Characteristics of cyclone climatology and variability in the Southern Ocean

A new climatology of cyclones in the Southern Ocean is generated by applying an automated cyclone detection and tracking algorithm （developed by Hodges at the Reading University） for an improved and relatively high- resolution European Centre for Medium-Range Weather Forecasts atmospheric reanalysis during 1979-2013. A validation shows that identified cyclone tracks are in good agreement with a available analyzed cyclone product. The climatological characteristics of the Southern Ocean cyclones are then analyzed, including track, number, density, intensity, deepening rate and explosive events. An analysis shows that the number of cyclones in the Southern Ocean has increased for 1979-2013, but only statistically significant in summer. Coincident with the circumpolar trough, a single high-density band of cyclones is observed in 55^-67~S, and cyclone density has generally increased in north of this band for 1979-2013, except summer. The intensity of up to 70% cyclones in the Southern Ocean is less than 980 hPa, and only a few cyclones with pressure less than 920 hPa are detected for 1979-2013. Further analysis shows that a high frequency of explosive cyclones is located in the band of 45^-55~S, and the Atlantic Ocean sector has much higher frequent occurrence of the explosive cyclones than that in the Pacific Ocean sector. Additionally, the relationship between cyclone activities in the Southern Ocean and the Southern Annular Mode is discussed.

Seasonal and inter-annual variations of Arctic cyclones and their linkage with Arctic sea ice and atmospheric teleconnections

The seasonal and inter-annual variations of Arctic cyclone are investigated. An automatic cyclone tracking algorithm developed by University of Reading was applied on the basis of European Center for Medium-range Weather Forecasts（ECMWF） ERA-interim mean sea level pressure field with 6 h interval for 34 a period. The maximum number of the Arctic cyclones is counted in winter, and the minimum is in spring not in summer.About 50% of Arctic cyclones in summer generated from south of 70°N, moving into the Arctic. The number of Arctic cyclones has large inter-annual and seasonal variabilities, but no significant linear trend is detected for the period 1979–2012. The spatial distribution and linear trends of the Arctic cyclones track density show that the cyclone activity extent is the widest in summer with significant increasing trend in CRU（central Russia）subregion, and the largest track density is in winter with decreasing trend in the same subregion. The linear regressions between the cyclone track density and large-scale indices for the same period and pre-period sea ice area indices show that Arctic cyclone activities are closely linked to large-scale atmospheric circulations, such as Arctic Oscillation（AO）, North Atlantic Oscillation（NAO） and Pacific-North American Pattern（PNA）. Moreover,the pre-period sea ice area is significantly associated with the cyclone activities in some regions.

AN APPLIED RESEARCH ON APPROACH OF DYADIC WAVELET TRANSFORM FOR REMOTE SENSING IMAGE EDGE DETECTION

In the edge detection of Remote Sensing (RS) image, the useful detail losing and the spurious edge often appear. To solve the problem, the authors uses the dyadic wavelet to detect the edge of surface features by combining the edge detecting with the multi-resolution analyzing of the wavelet transform. Via the dyadic wavelet decomposing, the RS image of a certain appropriate scale is obtained, and the edge data of the plane and the upright directions are respectively figured out, then the gradient vector module of the surface features is worked out. By tracing them, the authors get the edge data of the object, therefore build the RS image which obtains the checked edge. This method can depress the effect of noise and examine exactly the edge data of the object by rule and line. With an experiment of an RS image which obtains an airport, the authors certificate the feasibility of the application of dyadic wavelet in the object edge detection.

Motion Tracking with Fast Adaptive Background Subtraction

To extract and tr ack moving objects is usually one of the most important tasks of intelligent video surveillance systems. This paper presents a fast and adaptive background subtraction algorithm and the motion tracking process using this algorithm. The algorithm uses only luminance components of sampled image sequence pixels and models every pixel in a statistical model. The algorithm is characterized by its ability of real time detecting sudden lighting changes, and extracting and tracking motion objects faster. It is shown that our algorithm can be realized with lower time and space complexity and adjustable object detection error rate with comparison to other background subtraction algorithms. Making use of the algorithm, an indoor monitoring system is also worked out and the motion tracking process is presented in this paper. Experimental results testify the algorithm's good performances when used in an indoor monitoring system.

A Climatology of Extratropical Cyclones over East Asia During 1958-2001

A climatology of extratropical cyclones （ECs） over East Asia （20~ 75~N, 60^-160~E） is analyzed by applying an improved objective detection and tracking algorithm to the 4-time daily sea level pressure fields from the European Centre for Medium-range Weather Forecasts （ECMWF） reanalysis data. A total of 12914 EC processes for the period of 1958-2001 are identified, with an EC database integrated and EC activities reanalyzed using the objective algorithm. The results reveal that there are three major cyclogenesis regions： West Siberian Plain, Mongolia （to the south of Lake Baikal）, and the coastal region of East China; whereas significant cyclolysis regions are observed in Siberia north of 60~N, Northeast China, and Okhotsk Se^Northwest Pacific. It is found that the EC lifetime is largely 1 7 days while winter ECs have the shortest lifespan. The ECs are the weakest in summer among the four seasons. Strong ECs often appear in West Siberia, Northeast China, and Okhotsk Sea-Northwest Pacific. Statistical analysis based on k-means clustering has identified 6 dominating trajectories in the area south of 55~N and east of 80~E, among which 4 tracks have important impacts on weather/climate in China. ECs occurring in spring （summer） tend to travel the longest （shortest）. They move the fastest in winter, and the slowest in summer. In winter, cyclones move fast in Northeast China, some areas of the Yangtze-Huaihe River region, and the south of Japan, with speed greater than 15 m s-1. Explosively-deepening cyclones are found to occur frequently along the east coast of China, Japan, and Northwest Pacific, but very few storms occur over the inland area. Bombs prefer to occur in winter, spring, and autumn. Their annual number and intensity in 1990 and 1992 in East Asia （EA） are smaller and weaker than their counterparts in North America.