Baojie Fan

UDSFS: Unsupervised deep sparse feature selection

Abstract In this paper, we focus on unsupervised feature selection. As we have known, the combination of several feature units into a whole feature vector is broadly adopted for effective object representation, which may inevitably includes some irrelevant/redundant feature units or feature dimensions. Most of the traditional feature selection models can only select the feature dimensions without concerning the intrinsic relationship among different feature units. By taking into consideration the group sparsity of feature dimensions and feature units based on an l 2 , 1 minimization, we propose a new unsupervised feature selection model, unsupervised deep sparse feature selection (UDSFS) in this paper. In comparison with the state-of-the-arts, our UDSFS model can not only select the most discriminative feature units but also assign proper weight to the useful feature dimensions concurrently; moreover, the efficiency and robustness of our UDSFS can be also improved without extracting the discarded irrelevant feature units. For model optimization, we introduce an efficient iterative algorithm to solve the non-smooth, convex model and obtain a global optimization with the convergence rate as O ( 1 / K 2 ) (K is the iteration number). For the experiments, a new medical endoscopic image dataset, Abnormal Endoscopic Image Detection dataset (AEID), is built for evaluation; we also test our model using two public UCI datasets. Various experiments and comparisons with other state-of-the-arts justified the effectiveness and efficiency of our UDSFS model.

A robust template tracking algorithm with weighted active drift correction

In this paper, we propose a novel algorithm for object template tracking and its drift correction. It can prevent the tracking drift effectively, and save the time of an additional correction tracking. In our algorithm, the total energy function consists of two terms: the tracking term and the drift correction term. We minimize the total energy function synchronously for template tracking and weighted active drift correction. The minimization of the active drift correction term is achieved by the inverse compositional algorithm with a weighted L2 norm, which is incorporated into traditional affine image alignment (AIA) algorithm. Its weights can be adaptively updated for each template. For diminishing the accumulative error in tracking, we design a new template update strategy that chooses a new template with the lowest matching error. Finally, we will present various experimental results that validate our algorithm. These results also show that our algorithm achieves better performance than the inverse compositional algorithm for drift correction.

Discriminative multi-task objects tracking with active feature selection and drift correction

Abstract In this paper, we propose a discriminative multi-task objects tracking method with active feature selection and drift correction. The developed method formulates object tracking in a particle filter framework as multi-Task discriminative tracking. As opposed to generative methods that handle particles separately, the proposed method learns the representation of all the particles jointly and the corresponding coefficients are similar. The tracking algorithm starts from the active feature selection scheme, which adaptively chooses suitable number of discriminative features from the tracked target and background in the dynamic environment. Based on the selected feature space, the discriminative dictionary is constructed and updated dynamically. Only a few of them are used to represent all the particles at each frame. In other words, all the particles share the same dictionary templates and their representations are obtained jointly by discriminative multi-task learning. The particle that has the highest similarity with the dictionary templates is selected as the next tracked target state. This jointly sparsity and discriminative learning can exploit the relationship between particles and improve tracking performance. To alleviate the visual drift problem encountered in object tracking, a two-stage particle filtering algorithm is proposed to complete drift correction and exploit both the ground truth information of the first frame and observations obtained online from the current frame. Experimental evaluations on challenging sequences demonstrate the effectiveness, accuracy and robustness of the proposed tracker in comparison with state-of-the-art algorithms.

Speeded Up Low-Rank Online Metric Learning for Object Tracking

Visual object tracking can be considered as an online procedure to adaptively measure the foreground object similarity itself. However, many previous works usually adopt a fixed metric or offline metric learning to evaluate this dynamic process; even with some online metric learning (OML) trackers, their models often suffer from overfitting issues. To overcome these deficiencies, we propose a self-supervised tracking method that incorporates adaptive metric learning and semisupervised learning into a unified framework. For similarity measurement, we design a new OML model via low-rank constraint to handle overfitting. In particular, we employ the max norm instead of the trace norm used in our previous work. This not only maintains the low-rank property to overcome overfitting, but also reduces the computational complexity from O(n3) to O(n2), such that the new model is more suitable for object tracking. Moreover, by associating the information from stored training templates with unlabeled testing samples, a bilinear graph is defined accordingly to propagate the label of each sample. High-confidence samples are then collected for self-training the model and updating the templates concurrently to handle large scale. Experiments on various benchmark data sets and comparisons to several state-of-the-art methods demonstrate the effectiveness and efficiency of our algorithm.

The registration of UAV down-looking aerial images to satellite images with image entropy and edges

In this paper, we propose a novel and efficient image registration algorithm between high resolution satellite images and UAV down-looking aerial images. The algorithm is achieved by a composite deformable template matching. To overcome the limitations of environment changes and different sensors, and to remain image information, we fuse the image edge and entropy features as image representation. According to the altitude information of the UAV, we can get the scales of the down-looking aerial images relative to the satellite images. In the following, we perform an effective search strategy in the satellite images to find the best matching position. Different experimental results show that the proposed algorithm is effective and robust.

A novel color based object detection and localization algorithm

We propose a novel and robust color object detection and localization algorithm. Without a priori information about the number of objects, our method can detect all the objects with similar color feature in template. An improved histogram backprojection algorithm is used to find the object candidate regions. The weighted histogram intersection is used to verify the presence of objects. With the color feature in template, our method can detect and locate the objects accurately, get the number of objects, estimate their scales and orientations. Our experimental results on outdoor images obtained under different environments verify the effectiveness of our algorithm

Two‐step active contour method based on gradient flow

Purpose - Active contour can describe targets silhouette accurately and has been widely used in image segmentation and target tracking. Its main drawback is huge computation that is still not well resolved. The purpose of this paper is to optimize the evolving path of active contour, to reduce the computation cost and to make the evolution effectively. Design/methodology/approach - The contour-evolution process is separated into two steps: global translation and local deformation. The contour global translation and local deformation are realized by average and normal gradient flow of the evolving contour curve, respectively. Findings - When a contour is far away from the object to be segmented or tracked, the effective way of contour evolution is that it moves to the object without deformation first and then it deforms into the shape of the object when it moves on the object. Originality/value - The method presented in this paper can optimize the curve evolving path effectively without complicated calculation, such as rebuilding a new inner product, and its computation cost is largely reduced.

Active Contour Method with Separate Global Translation and Local Deformation

Active Contour can describe targets accurately and has been widely used in image segmentation and target tracking. Its main drawback is huge computation that is still not well resolved. In this paper, by analyzing curve gradient flow, the evolution of active contour is divided into two steps: global translation and local deformation. When the curve is far away from the object, the curve just does the translation motion. This method can optimize the curve evolving path and efficiency, and then the computation cost is largely reduced. Our experiments show that our method can segment and track object effectively.

Structured and weighted multi-task low rank tracker

Abstract Low rank subspace and multi-task learning have been introduced into object tracking to pursuit the accurate representation. However, many existing methods regularize all rank components equally, and shrink with the same threshold. In addition, these methods ignore the discriminative and structured information among tasks during the tracking. In this paper, we propose an online discriminative multi-task tracker with structured and weighted low rank regularization (ODMT-SL). Specifically, the total tracking task is accomplished by the combination of multiple subtasks, and each subtask corresponds to the trace of the image patch from the tracked object. In order to improve the flexibility of multi-task tracker, the weighted nuclear norm is introduced to adaptively assign different tracking importance on different rank components of multiple tasks. The geometric structure relationship among and inside candidates (or training samples) are mined to learn the collaborate representation, according to the discriminative subspace and optimal classifier. They are simultaneously learned and updated by minimizing the developed tracking model. The best candidate is selected by jointly evaluating the normalized metric. The proposed tracker is empirically compared with the state-of-the-art trackers on a large set of public video sequences. Both quantitative and qualitative comparisons demonstrate that the proposed algorithm performs well in terms of effectiveness, accuracy and robustness.

Consistent multi-layer subtask tracker via hyper-graph regularization

Develop the online multi-subtask learning framework for robust object tracking with novel task definition.The relationships among and inside candidates or training samples are mined by hyper-graph regularization.Simultaneously learn and update the adaptively discriminative subspace and classifier.Consistent multi-subtask tracker is a general model for most existing multi-task trackers. Most multi-task learning based trackers adopt similar task definition by assuming that all tasks share a common feature set, which cant cover the real situation well. In this paper, we define the subtasks from the novel perspective, and develop a structured and consistent multi-layer multi-subtask tracker with graph regularization. The tracking task is completed by the collaboration of multi-layer subtasks. Different subtasks correspond to the tracking of different parts in the target area. The correspondences of the subtasks among the adjacent frames are consistent and smooth. The proposed model introduces hyper-graph regularizer to preserve the global and local intrinsic geometrical structures among and inside target candidates or trained samples, and decomposes the representative matrix of the subtasks into two components: low-rank property captures the subtask relationship, group-sparse property identifies the outlier subtasks. Moreover, a collaborate metric scheme is developed to find the best candidate, by concerning both discrimination reliability and representation accuracy. We show that the proposed multi-layer multi-subtask learning based tracker is a general model, which accommodates most existing multi-task trackers with the respective merits. Encouraging experimental results on a large set of public video sequences justify the effectiveness and robustness of the proposed tracker, and achieve comparable performance against many state-of-the-art methods.