Object Target Tracking Research Articles

Bokeh-based target tracking for structural dynamic monitoring: A novel approach in variable lighting conditions

The application of Computer Vision (CV) techniques in Structural Dynamic Monitoring (SDM) eliminates the need for sensor installation and calibration, providing reliable monitoring results. However, conventional CV methods typically require time-intensive supervised pre-processing steps such as pattern extraction and machine learning. This paper introduces a novel Bokeh-effect-based target object tracking method for SDM that forgoes the need for such pre-processing, allowing for unsupervised, real-time, non-contact monitoring. This method, adaptable for diverse lighting conditions and employing consumer-grade cameras and computers, creates circles around target objects positioned beyond the depth of field of the camera lens, then detects and tracks the movement of these circles to facilitate structural dynamic monitoring. The proposed method is applied to a wind tunnel experiment on bridge girder sections to verify its accuracy and reliability, and the result is validated with the measurement from LASER transducers in the same wind tunnel experiment. This innovative approach advances efficient SDM by offering rapid response and real-time non-contact tracking capability with a high tolerance for target objects under regular and low-light conditions while avoiding the complexities linked with the pre-processing inherent in other supervised CV detection methods.

Real time object tracking using deep learning and OpenCV

This article summarizes the background and application fields of Open-Source Computer Vision Library (OpenCV) and deep learning and conducts research based on their object detection and tracking applications. The model algorithm selected in this article is the Convolutional Neural Networks (CNN) algorithm. CNN algorithm can be used for object detection in real-time scenarios. Moreover, the CNN algorithm has good credibility. The Python program developed based on CNN algorithm can effectively achieve real-time object tracking. The model shows good detection and tracking performance for trained targets and can be further applied in more specific scenarios in the future. Because deep learning can process large-scale data and recognize complex patterns, it can automatically learn and extract advanced features. Combining the two can achieve faster and more accurate detection and tracking of target objects. After having a large enough training sample size, it can detect and track the specified object more accurately. However, it is precisely due to the enormous sample size required for deep learning that there are still some difficulties in applying it to real-time object tracking. This article first discusses the use of supervised learning methods for deep understanding. When the input sample capacity is large enough, the machine can better reflect the real-time detection and tracking of the target object. But the time required will significantly increase. This issue still needs to be addressed in the subsequent research process.

Adaptive Target Tracking With Interacting Heterogeneous Motion Models

Multiple motion estimators such as an interacting multiple model (IMM) have been utilized to track target objects such as cars and pedestrians with diverse motion patterns. However, the standard IMM has limitations in combining motion models with different state definitions, so it cannot contain a complementary set of models that accurately work for all motion patterns. In this paper, we propose IMM-based adaptive target tracking with heterogeneous velocity representations and linear/curvilinear motion models. It can integrate four motion models with different state definitions and dimensions to be completely complimentary for all types of motions. We experimentally demonstrate the effectiveness of the proposed method with accuracy for various motion patterns using two types of datasets: synthetic datasets and real datasets. Experimental results show that the proposed method achieves the adaptive target tracking for diverse types of motion and also for various objects such as cars, pedestrians, and drones in the real world.

A ConvNext-Based and Feature Enhancement Anchor-Free Siamese Network for Visual Tracking

Existing anchor-based Siamese trackers rely on the anchor’s design to predict the scale and aspect ratio of the target. However, these methods introduce many hyperparameters, leading to computational redundancy. In this paper, to achieve outstanding network efficiency, we propose a ConvNext-based anchor-free Siamese tracking network (CAFSN), which employs an anchor-free design to increase network flexibility and versatility. In CAFSN, to obtain an appropriate backbone network, the state-of-the-art ConvNext network is applied to the visual tracking field for the first time by improving the network stride and receptive field. Moreover, A central confidence branch based on Euclidean distance is offered to suppress low-quality prediction frames in the classification prediction network of CAFSN for robust visual tracking. In particular, we discuss that the Siamese network cannot establish a complete identification model for the tracking target and similar objects, which negatively impacts network performance. We build a Fusion network consisting of crop and 3Dmaxpooling to better distinguish the targets and similar objects’ abilities. This module uses 3DMaxpooling to select the highest activation value to improve the difference between it and other similar objects. Crop unifies the dimensions of different features and reduces the amount of computation. Ablation experiments demonstrate that this module increased success rates by 1.7% and precision by 0.5%. We evaluate CAFSN on challenging benchmarks such as OTB100, UAV123, and GOT-10K, validating advanced performance in noise immunity and similar target identification with 58.44 FPS in real time.

Model-Agnostic Temporal Regularizer for Object Localization Using Motion Fields

Video analysis often requires locating and tracking target objects. In some applications, the localization system has access to the full video, which allows fine-grain motion information to be estimated. This paper proposes capturing this information through motion fields and using it to improve the localization results. The learned motion fields act as a model-agnostic temporal regularizer that can be used with any localization system based on keypoints. Unlike optical flow-based strategies, our motion fields are estimated from the model domain, based on the trajectories described by the object keypoints. Therefore, they are not affected by poor imaging conditions. The benefits of the proposed strategy are shown on three applications: 1) segmentation of cardiac magnetic resonance; 2) facial model alignment; and 3) vehicle tracking. In each case, combining popular localization methods with the proposed regularizer leads to improvement in overall accuracies and reduces gross errors.

Learning the Incremental Warp for 3D Vehicle Tracking in LiDAR Point Clouds

Object tracking from LiDAR point clouds, which are always incomplete, sparse, and unstructured, plays a crucial role in urban navigation. Some existing methods utilize a learned similarity network for locating the target, immensely limiting the advancements in tracking accuracy. In this study, we leveraged a powerful target discriminator and an accurate state estimator to robustly track target objects in challenging point cloud scenarios. Considering the complex nature of estimating the state, we extended the traditional Lucas and Kanade (LK) algorithm to 3D point cloud tracking. Specifically, we propose a state estimation subnetwork that aims to learn the incremental warp for updating the coarse target state. Moreover, to obtain a coarse state, we present a simple yet efficient discrimination subnetwork. It can project 3D shapes into a more discriminatory latent space by integrating the global feature into each point-wise feature. Experiments on KITTI and PandaSet datasets showed that compared with the most advanced of other methods, our proposed method can achieve significant improvements—in particular, up to 13.68% on KITTI.

TagAttention: Mobile Object Tracing With Zero Appearance Knowledge by Vision-RFID Fusion

We propose to study mobile object tracing, which allows a mobile system to report the shape, location, and trajectory of the mobile objects appearing in a video camera and identifies each of them with its cyber-identity (ID), even if the appearances of the objects are not known to the system. Existing tracking methods either cannot match objects with their cyber-IDs or rely on complex vision modules pre-learned from vast and well-annotated datasets including the appearances of the target objects, which may not exist in practice. We design and implement TagAttention, a vision-RFID fusion system that achieves mobile object tracing without the knowledge of the target object appearances and hence can be used in many applications that need to track arbitrary un-registered objects. TagAttention adopts the visual attention mechanism, through which RF signals can direct the visual system to detect and track target objects with unknown appearances. Experiments show TagAttention can actively discover, identify, and track the target objects while matching them with their cyber-IDs by using commercial sensing devices in complex environments with various multipath reflectors. It only requires around one second to detect and localize a new mobile target appearing in the video and keeps tracking it accurately over time.

FusionTalk: An IoT-Based Reconfigurable Object Identification System

Multisensor data fusion combines various information sources to produce a more accurate or complete description of the environment. This article studies an object identification (OID) system using multiple distributed cameras and Internet-of-Things (IoT) devices for better visualizability and reconfigurability. We first propose a data processing and fusing method to merge the detection results of different IoT devices and video cameras, in order to locate, identify, and track target objects in the monitored area. Then, we develop the FusionTalk system by integrating the data fusion techniques with IoTtalk, an IoT device management platform. FusionTalk is designed with flexibility, modularity, and expansibility, where cameras, IoT devices, and network applications are modularized and can be conveniently plugged in/out, reconfigured, and reused through graphical user interfaces. In FusionTalk, the scope and the target of surveillance can be flexibly configured and associated, and administrators can be warned and easily visualize the movement and behavior of specific objects. Our experimental evaluation of the data fusion algorithm in various scenarios shows an identification accuracy above 95%. Finally, theoretical and numerical analyses on the failure probability of pairing IoT devices with video objects by FusionTalk are presented. Extensive experiments are performed to demonstrate the pairing effectiveness in real-world scenarios with failure probability less than 0.01%.

Target Object Tracking with Portable Sensors in Wireless Sensor Network

Detection and tracking of moving target objects is one of the important problems of wireless sensor systems. In recent years, portability has become an important research area for the WSN community. Although it was never thought that the WSN arrangement was completely stable, portability was initially perceived as some of the difficulties that must be overcome, including network, inclusion and use of vitality. Target object dictates the accuracy of target objects with which the position of target objects can be estimated. This problem becomes a test, especially given the portability of the sensor and the target object, in which the directions of the sensor and the target object have to be captured. [1, 17] Through this review, we consider that we consider the question of following signals that release appropriate target object using compact sensors that depend on the social issue of the signal. Since the versatile movement of the lens is unclear, the portable sensor controller uses the approximation accumulated by a remote sensor so that the compact article indicates the time of appearance (TOA). [4] The portable sensor controller confirms the TOA estimation information of both compact target object and portable sensors to evaluate their areas before guiding the development of portable sensors to achieve the target object. We propose the calculation of approximation (min-max) to calculate the monitoring area, which can be effectively understood through quasi-different programming (SDP), and apply a cubic potential to the portable sensor path Can do. We measure the area of portable sensors and focus on each other to improve the following accuracy.[2, 5] We determine the characteristic relationship between several basic parameters of the frame and the trackingtarget object, including the thickness of the sensor, the range of detection, the portability of the sensor, and the target object. We examine the relationships and the ability to influence by multiple parameters of the framework and locate the base number of portable sensors that are required to maintain the tracking of target objects in an MSN. To further improve the execution of the framework, we propose a weighted monitoring calculation, using estimation information more efficiently.[3] Our entertainment results suggest that insufficiently insured calculations provide excellent follow-up that can be improved by requesting greatness with an equal number of sensors when contrasted and with the position of static sensors.

해상 장애물 추적을 위한 다중센서 융합 방법에 따른 성능비교

해상 장애물 추적을 위한 다중센서 융합 방법에 따른 성능비교

Learning transform-aware attentive network for object tracking

Existing trackers often decompose the task of visual tracking into multiple independent components, such as target appearance sampling, classifier learning, and target state inferring. In this paper, we present a transform-aware attentive tracking framework, which uses a deep attentive network to directly predict the target states via spatial transform parameters. During off-line training, the proposed network learns generic motion patterns of target objects from auxiliary large-scale videos. These leaned motion patterns are then applied to track target objects on test sequences. Built on the Spatial Transform Network (STN), the proposed attentive network is fully differentiable and can be trained in an end-to-end manner. Notably, we only fine-tune the pre-trained network in the initial frame. The proposed tracker requires neither online model update nor appearance sampling during the tracking process. Extensive experiments on OTB-2013, OTB-2015, VOT-2014 and UAV-123 datasets demonstrate the competitive performance of our method against state-of-the-art attentive tracking methods.

Automated and coupled services of advanced smart surveillance systems toward green IT: tracking, retrieval and digital evidence

Green Security is a new research field defining and investigating security solutions using an energy-aware perspective. Growing efforts and interests for an intelligent or smart surveillance system which is capable of automatically detecting and tracking target objects is in the spotlight in the security community. So far, these technologies are mainly aimed at single camera applications and are evolving with the demand for wide-area surveillance systems currently. However, the tracking techniques used on a single camera have limitations in providing effective crime prevention and countermeasures when an incident occurs since an object is not linked to other cameras. In addition, the use of multi-camera systems for wide-area surveillance not only produces large amounts of video data to be stored, but also have more technical requirements in the interrelation between cameras or server. It require a considerable amount of time, manpower and energy in multi-camera tracking and back-tracking of objects. Therefore, we propose the advanced smart surveillance system for wide-areas which is capable of the automated tracking and retrieval of target object and digital evidence-video collection. Furthermore, we considered the multiple-camera environment with non-overlapping views which includes more constraint conditions by various light changes. This system enables real-time object tracking, fast post-retrieval and selective digital evidence collection with economy of time, manpower, memory devices, and energy consumption. Also, this system is more energy-efficient since our schemes are organically connected to each other.

도로 상황인식을 위한 배경 및 로컬히스토그램 기반 객체 추적 기법

도로에서 발생되는 차량간 충돌사고, 교통 소통 상황, 보행자 사고 등 다양한 도로 상황을 모니터링 및 자동으로 인식하여 교통정보를 제공하거나 긴급구난 서비스를 제공하기 위한 다양한 기술이 개발되고 있다. 도로 모니터링을 통한 다양한 객체 추적 및 상황인식을 위해서는 잡음 및 겹침 등에 강인한 객체 추적 기술이 요구된다. 본 논문에서는 외부 환경에서 Background Subtraction, LK-Optical Flow, 지역 기반 히스토그램 특징의 결합을 통해 추적을 위한 몇 가지 추정 인자를 생성하고 이를 통해 변화가 있는 객체, 잡음에도 비교적 강인한 추적 방법을 제안한다. 구체적으로는 객체의 초기 움직임 정보를 검출하기 위해 옵티컬 플로우를 적용하여 컬러 정보 및 밝기 변화에 무관한 이동 정보를 측정한다. 측정된 정보를 기반으로 하여 지역 히스토그램 기반 검증을 통해 신뢰도를 판단한다. 신뢰도가 낮을 경우 배경 제거 정보와 지역 히스토그램 트래커의 정보를 혼합하여 새로운 위치를 추정한다. 실험을 통해 제안된 기법이 객체를 추적하고 있는 도중 나타날 수 있는 충돌, 새로운 특징의 등장, 크기 변화 상황에 강인하게 동작함을 제시한다. Compared with traditional video monitoring systems that provide a video-recording function as a main service, an intelligent video monitoring system is capable of extracting/tracking objects and detecting events such as car accidents, traffic congestion, pedestrian detection, and so on. Thus, the object tracking is an essential function for various intelligent video monitoring and surveillance systems. In this paper, we propose a background and local histogram-based object tracking approach for intelligent video monitoring systems. For robust object tracking in a live situation, the result of optical flow and local histogram verification are combined with the result of background subtraction. In the proposed approach, local histogram verification allows the system to track target objects more reliably when the local histogram of LK position is not similar to the previous histogram. Experimental results are provided to show the proposed tracking algorithm is robust in object occlusion and scale change situation.

Integrated region matching approach for tracking target objects in video sequences using HSV colour space

The object tracking is one of the important tasks in computer vision. In this work, an object tracking method is proposed, which use the properties of the HSV colour space for representing the object. The pixels in the objects are transformed either as true colour pixels or grey colour pixels and clusters are formed either as a true colour cluster or as a gray colour cluster. The adaptive K-means clustering is applied for clustering the video frames and a suitable post processing is carried out for obtaining information about all clusters. A suitable similarity measure has been proposed by which the object level similarity between the current frame and next frame are calculated. The performance of the proposed approach is evaluated using bench mark video sequences such as PETS, SPEVI and also some proprietary videos. The performance of proposed approach is found to be encouraging compared to recently proposed approaches.

A Cooperative Object Tracking System with Fuzzy-Based Adaptive Camera Selection

Abstract The intelligent environments, built upon many distributed sensors, are promising technology for ubiquitous interaction between robots and human beings. Especially, it is important to track target objects and get the positional information of them in such environments. This paper focuses on adapting camera selection for target tracking in multi-camera system. In this paper, a fuzzy automaton based camera selection method is introduced. In the proposed method, the camera selection decision is driven by fuzzy automaton based on the previously selected camera (previous camera state) and the tracking level of the object in each available camera. Simulations for evaluation of the proposed method and comparison with the previous method are presented. The results show that the proposed method is efficient for adaptive camera selection in multi-camera environment and helps easy construction of multi-camera placement. An actual multi-camera system with the proposed camera selection method was developed for checking tracking performance in the real environment. Experiments in the constructed system show that the proposed method suits well the camera selection task for tracking a moving object in the real intelligent environment.

Extending 3D Lucas–Kanade tracking with adaptive templates for head pose estimation

The Lucas–Kanade tracker (LKT) is a commonly used method to track target objects over 2D images. The key principle behind the object tracking of an LKT is to warp the object appearance so as to minimize the difference between the warped object’s appearance and a pre-stored template. Accordingly, the 2D pose of the tracked object in terms of translation, rotation, and scaling can be recovered from the warping. To extend the LKT for 3D pose estimation, a model-based 3D LKT assumes a 3D geometric model for the target object in the 3D space and tries to infer the 3D object motion by minimizing the difference between the projected 2D image of the 3D object and the pre-stored 2D image template. In this paper, we propose an extended model-based 3D LKT for estimating 3D head poses by tracking human heads on video sequences. In contrast to the original model-based 3D LKT, which uses a template with each pixel represented by a single intensity value, the proposed model-based 3D LKT exploits an adaptive template with each template pixel modeled by a continuously updated Gaussian distribution during head tracking. This probabilistic template modeling improves the tracker’s ability to handle temporal fluctuation of pixels caused by continuous environmental changes such as varying illumination and dynamic backgrounds. Due to the new probabilistic template modeling, we reformulate the head pose estimation as a maximum likelihood estimation problem, rather than the original difference minimization procedure. Based on the new formulation, an algorithm to estimate the best head pose is derived. The experimental results show that the proposed extended model-based 3D LKT achieves higher accuracy and reliability than the conventional one does. Particularly, the proposed LKT is very effective in handling varying illumination, which cannot be well handled in the original LKT.

Object tracking based on the modular active shape model

The active contour model (ACM) was proposed by Kass et al. in 1988 (M. Kass, A. Witkin, D. Terzopoulos, Snake: Active contour models, Int. J. Comput. Vision (1988) 321–331) for segmentation and tracking of target objects in image space. In their theory some kinds of energies were designed to extract the boundaries of targets by giving higher or lower values of snake energy to them. This concept has been revised and developed by a few engineers in the field of computer vision to make their active contour models have a prior shape information in their energy equations, more robust characteristics, adaptive abilities in the elastic parameters, and so on. In spite of these efforts on ACMs, some basic problems such as weakness to strong surrounding edges, the drift of snaxels due to the changes of illumination conditions and sensitivity to cluttered environments, are still substantial problems in the applications of ACM to real environments. This paper proposes the design and an implementation of a tracking scheme with a combination of the ACM and the active shape model (ASM), in which the point distribution models are systematically constructed on the projective point of view. In this paper there are three main contributions. Firstly, the combination of the ACM and the ASM is tested to implement a model-based visual tracking system. Secondly, a systematical approach is proposed to construct a few individual point distribution models (PDM) generated on the basis of the projection relation. Finally, the modular active shape model (MASM) is designed to integrate the results of the principal component analysis (PCA) on the individually generated PDMs. By this concept the independent ideas by designers on the shape variations of the target are naturally integrated into the MASM which is modularly constructed on the basis of individual PDMs. As a result, a model-based visual tracker, the MASM, is designed to overcome the problems of the ACM, while the ideas of designer are systematically integrated in this MASM to include the expected variations on the target.

３次元Ｍｕｌｔｉｖａｌｕｅｄ Ｗａｔｅｒｓｈｅｄ法を用いた動画像中の対象物体トラッキング

This paper extends the multivalued watershed algorithm, which has been proposed as a technique to segment, still pictures, into a 3-dimensional algorithm and applies it to the problem of tracking a target in image sequences. The proposed tracking technique first generates a 3-dimensional representation of target frames by arranging them temporally, and then applies the 3-dimensional multivalued watershed algorithm to the representation, with the target object segmented manually on a single selected frame. Further mechanisms, such as global motion compensation, are introduced to improve tracking precision. Experimental results presented later in this paper reveal that the proposed technique can accurately keep tracking target objects. Comparisons with conventional frame-by-frame tracking algorithms also illustrate the superiority of this approach.

A Video Information System for Sport Motion Analysis

In this paper, we present several important methods in a video information system for sport motion analysis. We first describe an efficient method for computing the position of one or more moving objects in a video sequence. We utilize computer vision routines for both target object tracking and feature value extraction. Then, we propose a multiple subsequence matching mechanism for comparing the difference in object motion between two video sequences. A mapping function, penalty table and the overall distance between two video sequences have been defined. We use the prominent feature points as our segmentation positions and separate the sequence into several subsequence groups. After an alignment process, we can find the mapping of the best matching and get an evaluation report between the two video sequences for motion analysis. The video information system described in this paper has successfully analysed several examples of sport motion, and is representative of techniques commonly used in practice.