Center Location Error Research Articles

Mutation based self regulating and self perception particle swarm optimization for efficient object tracking in a video

Object tracking is a pre-eminent task in many computer vision application.This paper addresses a realistic object tracking problem, considering single as well as multiple objects with different challenging perspectives. In view of this, a Mutation-based Self Regulating and self-Perception Particle Swarm Optimization (MSRSP-PSO) algorithm is proposed in order to explore the effective sampling strategy as well as to estimate and localize the optimum position of the target object. In conjunction with the optimization technique, an effective appearance model is also used for the reinforcement of tracker to achieve better accuracy. The appearance model is constructed with the joint distribution of the Local Binary Pattern (LBP)/Local Contrast Measure (LCM) based texture description and the Ohta color features. In pursuance of validating the efficacy of the proposed tracking algorithm, performance of other state-of-the-art methods such as, Particle Swarm Optimization (PSO) based tracking, Sequential Particle Swarm Optimization (SPSO) based tracking, and Adaptive Gaussian Particle Swarm Optimization (AGPSO) based tracking algorithms are compared with the proposed tracking algorithm. Both the quantitative and qualitative evaluation of the experiments, validate the effectiveness of the proposed technique.The performance measures include the convergence rate, center location error and the coverage test (PASCAL SCORE). Furthermore, the non-parametric statistical analysis using the Sign test and Wilcoxon signed rank test, is also figured out to validate the statistical superiority of the proposed MSRSP-PSO algorithm over others. The experimental results reveal that the proposed algorithm robustly track single as well as multiple similar objects with complex interaction.

Robust Visual Tracking via Semiadaptive Weighted Convolutional Features

In recent years, hierarchical features extracted from convolutional neural network (CNN) for robust visual tracking have been developed by several methods. As features from different layers characterize different information of target and are set fixed weighted parameters, the performance of traditional visual tracking methods based on CNN can be further improved. In this letter, we propose a novel online visual tracking method by using hierarchical convolutional features with semiadaptive weights. The responses from different layers are assessed by a novel loss function. It considers the log likelihood and the entropy term of each response. The layer with the lower loss value is set to a higher weight parameter and the layer with the higher loss value is set to a lower one. We further develop a target appearance pyramid to deal with the scale change and an online classifier to redetect targets in case of tracking failures. Extensive experiments on challenging videos demonstrate that our method can achieve better tracking results in terms of lower center location error and higher overlap rate.

Visual tracking using discriminative representation with ℓ2 regularization

In this paper, we propose a novel visual tracking method using a discriminative representation under a Bayesian framework. First, we exploit the histogram of gradient (HOG) to generate the texture features of the target templates and candidates. Second, we introduce a novel discriminative representation and l2-regularized least squares method to solve the proposed representation model. The proposed model has a closed-form solution and very high computational efficiency. Third, a novel likelihood function and an update scheme considering the occlusion factor are adopted to improve the tracking performance of our proposed method. Both qualitative and quantitative evaluations on 15 challenging video sequences demonstrate that our method can achieve more robust tracking results in terms of the overlap rate and center location error.

A Multiple Random Feature Extraction Algorithm for Image Object Tracking

This paper proposes an object-tracking algorithm with multiple randomly-generated features. We mainly improve the tracking performance which is sometimes good and sometimes bad in compressive tracking. In compressive tracking, the image features are generated by random projection. The resulting image features are affected by the random numbers so that the results of each execution are different. If the obvious features of the target are not captured, the tracker is likely to fail. Therefore the tracking results are inconsistent for each execution. The proposed algorithm uses a number of different image features to track, and chooses the best tracking result by measuring the similarity with the target model. It reduces the chances to determine the target location by the poor image features. In this paper, we use the Bhattacharyya coefficient to choose the best tracking result. The experimental results show that the proposed tracking algorithm can greatly reduce the tracking errors. The best performance improvements in terms of center location error, bounding box overlap ratio and success rate are from 63.62 pixels to 15.45 pixels, from 31.75% to 64.48% and from 38.51% to 82.58%, respectively.

The Fusion of Adaptive Color Attributes for Robust Compressive Tracking

In the process of target tracking, effective extraction and description of the target feature is very important, however, the existing methods mainly depend on the intensity image and ignore the color information. In this paper, an improved compressive tracking algorithm efficiently fusing adaptive color information is proposed for more robust tracking. At first, the CN color attribute is used to describe the color appearance of the target. In order to reduce the operation processing burden, the two most distinctive color components are extracted adaptively by PCA from 11-channel CN color space. In order to adapt to the scale change caused by motion, a scale-invariant normalized rectangular feature is proposed. Constant learning rate of naive Bayesian classifier parameters results in poor robustness of original compressive tracking algorithm, a novel non-linear parameter updating strategy based on double S-shape function is adopted to automatically adjust the learning rate for higher stability to interference. Finally, the scale-invariant appearance model combining the adaptive color information is integrated with particle filter frame to eliminate the negative effects such as scale change, occlusion and illumination change. Experimental results on testing sequences demonstrate the remarkable performance of our method. Compared with several well-known tracking algorithms, the average center location error is reduced to 9.88 while the suboptimal tracker is 23.9, the average overlap ratio increases by 7% points at least.

Visual tracking with conditionally adaptive multiple template update scheme for intricate videos

Tracking of moving objects in real-time scenes is a challenging research problem in computer vision. This is due to incessant live changes in the object features, background, occlusions, and illumination deviations occurring at different instances in the scene. With the objective of tracking visual objects in intricate videos, this paper presents a new color-independent tracking approach, the contributions of which are threefold. First, the illumination level of the sequences is maintained constant using fast discrete curvelet transform. Fisher information metric is calculated based on a cumulative score by comparing the template patches with a reference template at different timeframes. This metric is used for quantifying distances between the consecutive frame histogram distributions. Then, a novel iterative algorithm called conditionally adaptive multiple template update is proposed to regulate the object templates for handling dynamic occlusions effectively. The proposed method is evaluated on a set of extensive challenging benchmark datasets. Experimental results in terms of Center Location Error (CLE), Tracking Success Score (TSS), and Occlusion Success Score (OSS) show that the proposed method competes well with other relevant state-of-the-art tracking methods.

A novel approach of multi-stage tracking for precise localization of target in video sequences

Visual tracking methods are mostly based on single stage state estimation that limitedly caters to precise localization of target under dynamic environment such as occlusion, object deformation, rotation, scaling and cluttered background. In order to address these issues, we introduce a novel multi-stage coarse-to-fine tracking framework with quick adaptation to environment dynamics. The key idea of our work is to propose two-stage estimation of object state and to develop an adaptive fusion model. Coarse estimation of object state is achieved using optical flow and multiple fragments are generated around this approximation. Precise localization of object is obtained through evaluation of these fragments using three complementary cues. Adaptation of proposed tracker to dynamic environment changes is quick due to incorporation of context sensitive cue reliability, which encompass its direct application for development of expert system for video surveillance. In addition, proposed framework caters to object rotation and scaling through a random walk state model and rotation invariant features. The proposed tracker is evaluated over eight- benchmarked color video sequences and competitive results are obtained. As an average of the outcomes, we achieved mean center location error (in pixels) of 6.791 and F-measure of 0.78. Results demonstrate that proposed tracker not only outperforms various state-of-the-art trackers but also effectively caters to various dynamic environments.

Visual tracking via probabilistic collaborative representation

We present a probabilistic collaborative representation method under Bayesian framework for visual tracking. First, principal component analysis (PCA) basis vectors and squared templates are used to model the appearance of tracked object. Second, to decline the high complexity in traditional tracking methods via sparse representation, we demonstrate the mechanism of a probabilistic collaborative representation method and propose a fast method for computing the coefficients. Third, we introduce a PCA basis vectors update mechanism for the appearance change of the tracked object. Experiments on challenging videos demonstrate that our method can achieve better tracking results in terms of lower center location error and higher overlap rate.

Generalized ℓP-regularized representation for visual tracking

Collaborative representation has been successfully applied to visual tracking to powerfully use all the PCA basis vectors in the target subspace for object representation. However, collaborative representation always exists redundant features that may affect the performance of visual tracking. In this paper, a visual tracking algorithm is proposed by solving a generalized ℓp-regularized (0≤p≤1) problem within a Bayesian inference framework for the reduction of redundant features. To efficiently solve the minimization problem of ℓp-regularization, the Generalization of Soft-threshold (GST) operator is applied in the framework of iterative Accelerated Proximal Gradient (APG) approach. Moreover, the GST operator can also provide a unified framework to observe the effects of different sparsity for visual tracking. To show the feasibility of ℓp-regularizer, we choose the representative ℓ0.5-norm as the regularizer for the target coefficient and adjust the corresponding sparsity to be appropriate. Furthermore, we also introduce an extra ℓ0-regularized tracker to observe the effect of excessive sparsity in a unified framework. Experimental results on several challenging sequences demonstrate that the proposed tracker leads to a more favorable performance in terms of accuracy measures including the overlap ratio and center location error, respectively.

Efficient compressive sensing tracking via mixed classifier decision

Recent years have witnessed successful use of tracking-by-detection methods, with a number of promising results being achieved. Most of these algorithms use a sliding window to collect samples and then employ these samples to train and update the classifiers. They also use an updated classifier to establish the appearance model and they take the maximum response value of the classifier as the location of the target within a fixed radius. Compressive Tracking (CT) is a novel tracking-by-detection algorithm that updates the appearance model in a compressed domain. However, the conventional CT algorithm uses a single classifier to detect the target, and if the selected region drifts, the classifier may become inaccurate. Furthermore, the CT algorithm updates the classifier parameters with a constant learning rate. Therefore, if the target is completely occluded for an extended period, the classifier will instead learn the features of the covered object and the target will ultimately be lost. To overcome these problems, we present a compressive sensing tracking algorithm using mixed classifier decision. The main improvements in our algorithm are that it adopts mixed classifiers to locate the target and it applies a dynamic learning rate to update the appearance model. An experimental comparison with state-of-the-art algorithms on eight benchmark video sequences in complicated situations shows that the proposed algorithm achieves the best performance with 12 pixels on the average center location error and 66.82\\% on the average overlap score.

Improved Circulant Structure with Kernel Tracker using Correlation Kernel based on Wavelet Feature

Visual object tracking has many applications related to computer vision. Recently, correlation filter based trackers have been ranked as the highest performers in this field. However, handling chronic problems such as occlusion, deformation, and scale variations is difficult with such trackers. These problems are solved by many other researches that employ other features and improve an appearance update. In this paper, we propose an improved CSK (Circulant Structure with Kernel) tracker using object feature decomposition in the wavelet domain. Specifically, a newly created correlation kernel is generated from different filter-banks reflecting the visual properties of a given object, and it is stable and robust to the environmental variations. Experimental results demonstrate that the proposed scheme outperforms the conventional CSK tracker in terms of center location error by 59% on an average for 100 sequences. Therefore, we believe that the proposed tracker can be useful for robust object tracking in occlusion.

An Adaptive Approach for Validation in Visual Object Tracking

Abstract One of the major goals in the field of computer vision is to enable computers to replicate the basic functions of human vision such as motion perception and scene understanding. To achieve the goal of intelligent motion perception, much effort has been spent on visual object tracking. Research interest in visual object tracking comes from the fact that it has a wide range of real-world applications. The uncertainty of validating unpredictable features in object tracking is a challenging task in visual object tracking with occlusion and large appearance variation. To address this uncertainty, we propose an adaptive approach which uses updating model based on the occlusion and distortion parameters. In case of occlusion or large appearance variation, the proposed method uses backward model validation where it updates the invalid appearance and then validates the target feature model. If the target feature did not undergo any kind of clutter or distortions, it simply validates and then updates the appearance model using forward feature validation. The experimental results obtained from this adaptive approach demonstrate effectiveness in terms of OR (Overlap Rate) and Center Location Error, compared with other relevant existing algorithms.

Robust Visual Correlation Tracking

Recent years have seen greater interests in the tracking-by-detection methods in the visual object tracking, because of their excellent tracking performance. But most existing methods fix the scale which makes the trackers unreliable to handle large scale variations in complex scenes. In this paper, we decompose the tracking into target translation and scale prediction. We adopt a scale estimation approach based on the tracking-by-detection framework, develop a new model update scheme, and present a robust correlation tracking algorithm with discriminative correlation filters. The approach works by learning the translation and scale correlation filters. We obtain the target translation and scale by finding the maximum output response of the learned correlation filters and then online update the target models. Extensive experiments results on 12 challenging benchmark sequences show that the proposed tracking approach reduces the average center location error (CLE) by 6.8 pixels, significantly improves the performance by 17.5% in the average success rate (SR) and by 5.4% in the average distance precision (DP) compared to the second best one of the other five excellent existing tracking algorithms, and is robust to appearance variations introduced by scale variations, pose variations, illumination changes, partial occlusion, fast motion, rotation, and background clutter.

Object tracking based on background-weighted spatial histogram

In order to improve the feature robustness of the traditional object tracking algorithm,an improved algorithm based on background-weighted spatial histogram was proposed.It combines the background-weighted histogram and spatial histogram and exploits the advantages of both features,which distinctively improves the feature robustness.The proposed algorithm was implemented under the framework of Mean Shift tracking and obtained a stable tracking.Experimental results indicate that the proposed algorithm can well track the object and obtains at least 80% decreasing of the mean center location error than that of the traditional algorithms.

1/32° real-time global ocean prediction and value-added over 1/16° resolution

A 1/32° global ocean nowcast/forecast system has been developed by the Naval Research Laboratory at the Stennis Space Center. It started running at the Naval Oceanographic Office in near real-time on 1 Nov. 2003 and has been running daily in real-time since 1 Mar. 2005. It became an operational system on 6 March 2006, replacing the existing 1/16° system which ceased operation on 12 March 2006. Both systems use the NRL Layered Ocean Model (NLOM) with assimilation of sea surface height from satellite altimeters and sea surface temperature from multi-channel satellite infrared radiometers. Real-time and archived results are available online at http://www.ocean.nrlssc.navy.mil/global_nlom. The 1/32° system has improvements over the earlier system that can be grouped into two categories: (1) better resolution and representation of dynamical processes and (2) design modifications. The design modifications are the result of accrued knowledge since the development of the earlier 1/16° system. The improved horizontal resolution of the 1/32° system has significant dynamical benefits which increase the ability of the model to accurately nowcast and skillfully forecast. At the finer resolution, current pathways and their transports become more accurate, the sea surface height (SSH) variability increases and becomes more realistic and even the global ocean circulation experiences some changes (including inter-basin exchange). These improvements make the 1/32° system a better dynamical interpolator of assimilated satellite altimeter track data, using a one-day model forecast as the first guess. The result is quantitatively more accurate nowcasts, as is illustrated by several model-data comparisons. Based on comparisons with ocean color imagery in the northwestern Arabian Sea and the Gulf of Oman, the 1/32° system has even demonstrated the ability to map small eddies, 25–75 km in diameter, with 70% reliability and a median eddy center location error of 22.5 km, a surprising and unanticipated result from assimilation of altimeter track data. For all of the eddies (50% small eddies), the reliability was 80% and the median eddy center location error was 29 km. The 1/32° system also exhibits improved forecast skill in relation to the 1/16° system. This is due to ( a) a more accurate initial condition for the forecast and ( b) better resolution and representation of critical dynamical processes (such as upper ocean – topographic coupling via mesoscale flow instabilities) which allow the model to more accurately evolve these features in time while running in forecast mode (forecast atmospheric forcing for the first 5 days, then gradually reverting toward climatology for the remainder of the 30-day forecast period). At 1/32° resolution, forecast SSH generally compares better with unassimilated observations and the anomaly correlation of the forecast SSH exceeds that from persistence by a larger amount than found in the 1/16° system.

Propagation of the hip joint centre location error to the estimate of femur vs pelvis orientation using a constrained or an unconstrained approach

To estimate hip joint angles during selected motor tasks using stereophotogrammetric data, it is necessary to determine the hip joint centre position. The question is whether the errors affecting that determination propagate less to the angles estimates when a three degrees of freedom (DOFs) constraint (spherical hinge) is used between femur and pelvis, rather than when the two bones are assumed to be unconstrained (six DOFs). An analytical relationship between the hip joint centre location error and the joint angle error was obtained limited to the planar case. In the 3-D case, a similar relationship was obtained using a simulation approach based on experimental data. The joint angle patterns resulted in a larger distortion using a constrained approach, especially when wider rotations occur. The range of motion of the hip flexion–extension, obtained simulating different location errors and without taking into account soft tissue artefacts, varied approximately 7 deg using a constrained approach and up to 1 deg when calculated with an unconstrained approach. Thus, the unconstrained approach should be preferred even though its estimated three linear DOFs most unlikely carry meaningful information.

Concatenated Approach to Trajectory Optimization

the covariance simulationincorporatesthe equations developedpreviously.Inasmuchasthe Tomahawkmissileis being considered,the terrain scene is a terminal area digital scene matching and area correlation(DSMAC) scene. Shown in Fig. 2 is a typical CEP plot from the covariance simulation, which illustrates that the highly accurate positional solution obtained from the relative GPS correction vector scheme degrades only minimally between the DSMAC scene and the target. For target penetration considerations, the results presented here incorporate a terminal trajectory with a high-angle dive to impact the target. At the target, the navigational error contributes 2.6%, the steering error 10.2%, the scene center location error 20.9%, and the relative target location error 66.3%, which is consistent with the earlier discussion regarding relative error magnitudes. An examination of Fig. 2 also reveals that the correction vector scheme affords a substantial improvement over the situation where the Kalman e lter accepts a high-quality DSMAC positional update and, subsequently, the missile e ies in a free inertial manner to the target (the mode of operation for the current Block III Tomahawk missile ). Because improvements in satellite imaging technology are expected to permit larger DSMAC scene to target separations while maintaining the same scene/target relative error, free inertial navigation to the target will clearly become less desirable, and relative GPS schemes such as the one proposed herein more desirable. Summary