Visual Tracking System Research Articles

Vision System-Based Design and Assessment of a Novel Shoulder Joint Mechanism for an Enhanced Workspace Upper Limb Exoskeleton.

Exoskeletons arise as the common ground between robotics and biomechanics, where rehabilitation is the main field in which these two disciplines find cohesion. One of the most relevant challenges in upper limb exoskeleton design relies in the high complexity of the human shoulder, where current devices implement elaborate systems only to emulate the drifting center of rotation of the shoulder joint. This paper proposes the use of 3D scanning vision technologies to ease the design process and its implementation on a variety of subjects, while a motion tracking system based on vision technologies is applied to assess the exoskeleton reachable workspace compared with an asymptomatic subject. Furthermore, the anatomic fitting index is proposed, which compares the anatomic workspace of the user with the exoskeleton workspace and provides insight into its features. This work proposes an exoskeleton architecture that considers the clavicle motion over the coronal plane whose workspace is determined by substituting the direct kinematics model with the dimensional parameters of the user. Simulations and numerical examples are used to validate the analytical results and to conciliate the experimental results provided by the vision tracking system.

High-precision quality assurance of robotic couches with six degrees of freedom

A robotic couch capable of six degrees of freedom (6-DoF) of motion was introduced for state-of-the-art radiation therapy. Patient treatment requires precise quality assurance (QA) of 6-DoF. Unfortunately, conventional methods do not provide the requisite accuracy and precision. Therefore, we developed a high-precision automated QA system using a visual tracking system (VTS). The VTS comprises four motion-sensing cameras, a cube with infrared reflective markers. To acquire data in treatment room coordinates, a transformation matrix from VTS coordinates to treatment room coordinates was determined.The mean error and standard deviation of linear and rotational motions, as well as couch sagging were analyzed from continuously acquired images in the moving couch. The accuracy of VTS was 0.024 mm deviation for the sinusoidal motion, and the accuracy of the transformation matrix was 0.02 mm. In a cross-comparison, the difference between Laser Tracker (FARO) measurements was 0.14 ± 0.12 mm for translation and 0.032 ± 0.026° on average for yaw rotation. The new system provides QA of yaw, pitch and roll motion as well as sagging of the couch and sub-millimeter/degree accuracy together with precision.

Underwater Platform for Intelligent Robotics and its Application in Two Visual Tracking Systems

A hovering-type autonomous underwater vehicle (AUV) capable of cruising at low altitudes and observing the seafloor using only mounted sensors and payloads was developed for sea-creature survey. The AUV has a local area network (LAN) interface for an additional payload that can acquire navigation data from the AUV and transmit the target value to the AUV. In the handling process of the state flow of an AUV, additional payloads can control the AUV position using the transmitted target value without checking the AUV condition. In the handling process of the state flow of an AUV, additional payloads can control the AUV position using the transmitted target value without checking the AUV condition. In this research, water tank tests and sea trials were performed using an AUV equipped with a visual tracking system developed in other laboratories. The experimental results proved that additional payload can control the AUV position with a standard deviation of 0.1 m.

Research of a non-linearity control algorithm for UAV target tracking based on fuzzy logic systems

Target tracking is one of the most widely used applications of UAVs and visual based tracking is the main method for non-cooperative tracking. In non-cooperative tracking missions, the UAV circles around the target and the gimbal rotates to keep the optic axis of the onboard camera pointing to the target. Compared with the biaxial gimbal, the single-axis gimbal system consists of only one torque motor, forming a great lightweight sensor suite, which also increases the requirement of the control accuracy. In the process of tracking moving target using a UAV with single-axis gimbal, there are still several challenges: (1) The completion of high precision and reliability of moving target tracking using UAV with only single-axis gimbal. (2) The non-linearity and uncertainty in the UAV system, where the non-linearity exists in the control of altitude, heading angle and roll angle. (3) The uncertainty in the expected turning radius of trajectory of the UAV when tracking a moving target without knowing its motion state. This paper proposes a vision-based fuzzy controller for a target tracking system consists of a fixed-wing UAV with single-axis gimbal. In this research, the innovations are described as follows: (1) A control algorithm is proposed for visual target tracking system consists of fixed-wing UAV with single-axis gimbal, which is able to guide the UAV to complete tracking task precisely and reliably. (2) Generation of roll command and heading command is immediately based on the information obtained from the images, skipping the step of calculating the velocity and position of the target, which can avoid unnecessary errors. (3) To deal with the non-linearities and uncertainties in the tracking system, seven fuzzy controllers are used to keep UAV circling around the target stably. (4) Flight tests are accomplished to verify the algorithm. Simulation results show that the maximum angle offset of the camera’s optic axis is 0.04°, and the angle offsets can be kept in the range of 5° in the further flight test, which shows that the algorithm is able to accomplish the task of tracking a moving target successfully.

Research on eye localization method based on adaptive correlation filter

Eye location is one of the most important steps in the face recognition and visual tracking system. This article combines the correlation filter with integral projection to detect eye position precisely, and puts forward two improvements in the training and test phases. Firstly, the adaptive synthetic correlation filter are rotated within the angle range of –0.2, –0.1, 0, 0.1, 0.2, and the corresponding location of the maximum grey value was served as an initial anchor point. Secondly, within the anchor point 5 × 5 neighbourhood, grey-level integral projection is implemented in the horizontal and vertical direction respectively. Finally, the ultimate goal orientation chooses the position of minimum integration. Compared with the average synthetic exact filter, minimum output sum of squared error filter and adaptive synthetic correlation filter, the experimental results show the proposed improvements are superior to the general algorithm in the accuracy and stability.

Research on eye localisation method based on adaptive correlation filter

Eye location is one of the most important steps in the face recognition and visual tracking system. This article combines the correlation filter with integral projection to detect eye position precisely, and puts forward two improvements in the training and test phases. Firstly, the adaptive synthetic correlation filter are rotated within the angle range of –0.2, –0.1, 0, 0.1, 0.2, and the corresponding location of the maximum grey value was served as an initial anchor point. Secondly, within the anchor point 5 × 5 neighbourhood, grey-level integral projection is implemented in the horizontal and vertical direction respectively. Finally, the ultimate goal orientation chooses the position of minimum integration. Compared with the average synthetic exact filter, minimum output sum of squared error filter and adaptive synthetic correlation filter, the experimental results show the proposed improvements are superior to the general algorithm in the accuracy and stability.

Technical, tactical and movement analysis of men's professional tennis on hard courts.

There is very little information in tennis relating performance to both the movement of the players and the effectiveness of their shots. This study aimed to identify differences between winners and losers in the volume and intensity of their movements, as well as in the effectiveness of their shots. Matches (N.=8) at the 2011 ATP 500 tournament in Valencia (Spain) were recorded and analyzed using SAGIT, a computerized vision-tracking system. Data were split into points for movement analysis, and into games for stroke analysis. Players winning points used more offensive strategies and spent more time in the offensive zone than point losers. They also forced point losers to cover more distance at a higher speed. Game winners hit more winning shots, made fewer unforced errors, and were more effective than game losers. These results suggest that winner players controlled the game by playing more offensively and forcing losers to play in defensive positions, move greater distances at a faster speed, and make more errors.

Distributed extended Kalman filter with nonlinear consensus estimate

This paper is concerned with the distributed filtering problem for discrete-time nonlinear systems over a sensor network. In contrast with the distributed filters with linear consensus estimate, a distributed extended Kalman filter (EKF) is developed with nonlinear consensus estimate. Specifically, a new nonlinear consensus protocol with polynomial form is proposed to generate the consensus estimate. By using the variance-constrained approach, the Kalman gain matrix is determined for each node to guarantee an optimized upper bound on the state estimation error covariance despite consensus terms and linearization errors. It is shown that the Kalman gain matrix can be derived by solving two Riccati-like difference equations. The effectiveness of the proposed filter is evaluated on an indoor localization of a mobile robot with visual tracking systems.

Design and development of an on-board autonomous visual tracking system for unmanned aerial vehicles

This paper proposes the design and development of an on-board autonomous visual tracking system (AVTS) for unmanned aerial vehicles (UAV). A prototype of the proposed system has been implemented in MATLAB/ Simulink for simulation purposes. The proposed system contains GPS/INS sensors, a gimbaled camera, a multi-level autonomous visual tracking algorithm, a ground stationary target (GST) or ground moving target (GMT) state estimator, a camera control algorithm, a UAV guidance algorithm, and an autopilot. The on-board multi-level autonomous visual tracking algorithm acquires the video frames from the on-board camera and calculates the GMT pixel position in the video frame. The on-board GMT state estimator receives the GMT pixel position from the multi-level autonomous visual tracking algorithm and estimates the current position and velocity of the GMT with respect to the UAV. The on-board non-linear UAV guidance law computes the UAV heading velocity rates and sends them to the autopilot to steer the UAV in the desired path. The on-board camera control law computes the control command and sends it to the camera's gimbal controller to keep the GMT in the camera's field of view. The UAV guidance law and camera control law have been integrated for continuous tracking of the GMT. The on-board autopilot is used for controlling the UAV trajectory. The simulation of the proposed system was tested with a flight simulator and the UAV's reaction to the GMT was observed. The simulated results prove that the proposed system tracks a GST or GMT effectively.

심층신경망 기반 우주파편 영상 추적시스템 인식모델에 대한 연구

심층신경망 기반 우주파편 영상 추적시스템 인식모델에 대한 연구

이동 로봇용 안드로이드 기반 비전 추적 시스템

이동 로봇용 안드로이드 기반 비전 추적 시스템

Real-time multi-scale parallel compressive tracking

Robust visual tracking is a challenging problem because the appearance of a target may rapidly change due to significant variations in the object’s motion and the surrounding illumination. In this paper, a novel robust visual tracking algorithm is proposed based on an existing compressive tracking method. The proposed algorithm adopts multiple naive Bayes classifiers, each trained under a different scale condition, to realize online parallel multi-scale classification. Further, each classifier was initialized by randomly generating different types of Haar-like features. By doing so, the robustness of the feature classification can be improved to obtain more accurate tracking results. To enhance the real-time performance of the visual tracking system, the formula of the naive Bayes classifier is studied and simplified to speed up the processing speed of parallel multi-scale feature classification. After acceleration via formula simplification and parallel implementation, the proposed visual tracking algorithm can reach a tracking performance of approximately 45 frames per second (fps) when dealing with images of 642 × 352 pixels on a popular Intel Core i5-3230M platform. The experimental results show that the proposed algorithm outperforms state-of-the-art visual tracking methods on challenging videos in terms of success rate, tracking accuracy, and visual comparison.

A Real-time Visual Tracking System with Delivery Robot

In this paper, we proposed a real-time visual tracking system for delivery robot. The RGB-D data are collected for environmental perception. The critical step is the tracking algorithm we utilized, this tracking algorithm must be robust, which can easily detect and locate the target, and adapt many complicated environment without loss the target. And moreover, this tracking algorithm must meet the high demand of real-time capability to fulfil the task. In the system designing, a FFT based tracking algorithm is used and the experimental results showed that this algorithm is both robust and efficient, thus is very appropriate for the luggage delivery robotic system.

Visual Detection and Tracking System for a Spherical Amphibious Robot.

With the goal of supporting close-range observation tasks of a spherical amphibious robot, such as ecological observations and intelligent surveillance, a moving target detection and tracking system was designed and implemented in this study. Given the restrictions presented by the amphibious environment and the small-sized spherical amphibious robot, an industrial camera and vision algorithms using adaptive appearance models were adopted to construct the proposed system. To handle the problem of light scattering and absorption in the underwater environment, the multi-scale retinex with color restoration algorithm was used for image enhancement. Given the environmental disturbances in practical amphibious scenarios, the Gaussian mixture model was used to detect moving targets entering the field of view of the robot. A fast compressive tracker with a Kalman prediction mechanism was used to track the specified target. Considering the limited load space and the unique mechanical structure of the robot, the proposed vision system was fabricated with a low power system-on-chip using an asymmetric and heterogeneous computing architecture. Experimental results confirmed the validity and high efficiency of the proposed system. The design presented in this paper is able to meet future demands of spherical amphibious robots in biological monitoring and multi-robot cooperation.

High-Speed Object Tracking with Its Application in Golf Playing

This paper presents a new high-speed object tracking algorithm and applies it in tracking high-speed golf balls. Several challenges are encountered in tracking this kind of small-size and high-speed objects; these challenges include: (1) the motion speed (up to 200km/h) of the object is extremely fast that the image captured by the camera can become easily blurred; (2) the size of the objects is small and there is not enough texture feature on the surface of the objects; (3) high-speed objects typically move in outdoor environments, which have complex backgrounds and the light conditions; and (4) a balance between robustness and real-time efficiency is difficult to achieve. There is no existing method can directly solve all these problems. To address these problems, a novel real-time visual tracking algorithm and system are developed that can realize high-speed small object tracking and trajectory predication. The frame difference method was employed to segment the moving objects from the background. The motion blurred object can be detected by applying the proposed object recognition method using multi-features (geometry features and object motion features). Different from the state of the art tracking algorithms, the motion blurring feature was also considered as a feature of the high-speed object, and the tracking robustness of the proposed algorithm does not rely on the texture feature of the tracking objects. To realize real-time efficiency, a region of interest predicting and updating method was proposed that can shrink the processing area of the frames and achieve a high frame rate. Combining visual servoing technology and aerodynamics, the trajectory of the high-speed object can be predicted. The high-speed golf ball tracking experiments were conducted to demonstrate the tracking efficiency and robustness of the algorithm.

An experimental characterization of human falling down

Abstract. This paper presents results of an experimental investigation on the falling down of the human body in order to identify significant characteristics and parameters. A specific lab layout has been settled up with vision tracking system and suitable sensors to monitor information on trajectories, impact force and acceleration during the falling with elaboration procedures that make fairly easy to track and interpret the motion characteristics. We focus on the more often falling mode: forward and backward falling Tests are discussed with results from lab tests that give both behavior and values of the biomechanics of falling down of the human body. Possible protection strategies for falling based on the proposed research are talked about at the last.

A novel foreground region analysis using NCP-DBP texture pattern for robust visual tracking

Robust visual tracking is the important stage in the computer vision applications such as robotics, man-free control systems, and the visual surveillance. Accurate motion states estimation and the target representation in visual tracking system are based on the appearances of the target. The factor affects the learning of target representation is the accumulated error due to the pose, illumination changes, and the uneven background. The presence of dynamic background and the shadowing effects causes the visual drift and destructive information. Besides, the misclassification of target region induces the false detection of moving objects. The K-means and Fuzzy-C-means clustering algorithms are available to segment the foreground/background and suppress the shadow region on the basis of the non-changing background of the surveillance area. This paper proposes the novel background normalization technique with textural pattern analysis to suppress the shadow region. The Neighborhood Chain Prediction (NCP) algorithm is used to cluster the uneven background and the Differential Boundary Pattern (DBP) extracts the texture of the video frame to suppress the shadow pixels present in the frame. The lower intensity estimation and the prediction of the area around the lower intensity in proposed work enhance the pixels for shadow removal. The shadow-free frame split up into several grids and the histograms of features are extracted from the grid formatted frame. Finally, the Machine Level Classification (MLC) finds the matching grid corresponds to the tracking region and provides the binary labeling to separate the background and foreground. The proposed DBP-based visual tracking system is high robustness over the sudden illumination changes and the dynamic background due to the texture pattern analysis. The comparison of proposed NCP-DBP combination with the existing segmentation techniques regarding the accuracy, precision, recall, F-measure, success and error rate assured the effectiveness in visual tracking applications.

Visual object tracking via enhanced structural correlation filter

In this study, we aim to build a robust correlation-based visual object tracking system. The function of traditional correlation filters for visual tracking is to search the most likely position of target by circularly shifting the search image patch. However, the search image patch needs to be large enough to cover both the object and background, which results in the algorithm being sensitive to changes in background. To alleviate this problem, we first propose an efficient object-surrounding histogram model to suppress the background. In this model, we build a Bayes classifier based on the initial given object, and we then apply it to each pixel in subsequent frames. With this model, the original image can be enhanced in order to eliminate the impact of circular shifting. Moreover, we develop a structural correlation filter that consists of both holistic and local object parts. The multiple object parts are adaptively weighted and further aggregated to predict the relative motion from the last frame. We conduct extensive experiments on frequently used benchmarks with 51 video sequences. The experimental results show that the proposed algorithm achieves outstanding performance, especially in terms of heavy occlusion and severe deformation.

Distributed consensus extended Kalman filter: a variance‐constrained approach

This study is concerned with the distributed state estimation problem for non-linear systems over sensor networks. By using the strategy of consensus on prior estimates, a distributed extended Kalman filter (EKF) is developed for each node to guarantee an optimised upper bound on the state estimation error covariance despite consensus terms and linearisation errors. The Kalman gain matrix is derived for each node by solving two Riccati-like difference equations. It is shown that the estimation error is bounded in mean square under certain conditions. The effectiveness of the proposed filter is evaluated on an indoor localisation of a mobile robot with visual tracking systems.

Correction of hysteretic respiratory motion in SPECT myocardial perfusion imaging: Simulation and patient studies.

Amplitude-based respiratory gating is known to capture the extent of respiratory motion (RM) accurately but results in residual motion in the presence of respiratory hysteresis. In our previous study, we proposed and developed a novel approach to account for respiratory hysteresis by applying the Bouc-Wen (BW) model of hysteresis to external surrogate signals of anterior/posterior motion of the abdomen and chest with respiration. In this work, using simulated and clinical SPECT myocardial perfusion imaging (MPI) studies, we investigate the effects of respiratory hysteresis and evaluate the benefit of correcting it using the proposed BW model in comparison with the abdomen signal typically employed clinically. The MRI navigator data acquired in free-breathing human volunteers were used in the specially modified 4D NCAT phantoms to allow simulating three types of respiratory patterns: monotonic, mild hysteresis, and strong hysteresis with normal myocardial uptake, and perfusion defects in the anterior, lateral, inferior, and septal locations of the mid-ventricular wall. Clinical scans were performed using a Tc-99m sestamibi MPI protocol while recording respiratory signals from thoracic and abdomen regions using a visual tracking system (VTS). The performance of the correction using the respiratory signals was assessed through polar map analysis in phantom and 10 clinical studies selected on the basis of having substantial RM. In phantom studies, simulations illustrating normal myocardial uptake showed significant differences (P<0.001) in the uniformity of the polar maps between the RM uncorrected and corrected. No significant differences were seen in the polar map uniformity across the RM corrections. Studies simulating perfusion defects showed significantly decreased errors (P<0.001) in defect severity and extent for the RM corrected compared to the uncorrected. Only for the strong hysteretic pattern, there was a significant difference (P<0.001) among the RM corrections. The errors in defect severity and extent for the RM correction using abdomen signal were significantly higher compared to that of the BW (severity=-4.0%, P<0.001; extent=-65.4%, P<0.01) and chest (severity=-4.1%, P<0.001; extent=-52.5%, P<0.01) signals. In clinical studies, the quantitative analysis of the polar maps demonstrated qualitative and quantitative but not statistically significant differences (P=0.73) between the correction methods that used the BW signal and the abdominal signal. This study shows that hysteresis in respiration affects the extent of residual motion left in the RM-binned data, which can impact wall uniformity and the visualization of defects. Thus, there appears to be the potential for improved accuracy in reconstruction in the presence of hysteretic RM with the BW model method providing a possible step in the direction of improvement.