Search

Visual tracking is fundamental in computer vision tasks. The Siamese-based trackers have shown surprising effectiveness in recent years. However, two points have been neglected: firstly, few of them focus on fusing the image level and semantic level features in neural networks, which usually resulting in tracking failure when differentiating the target from other distractors of the same class. Secondly, the robustness of the previous redetection scheme is limited by simply expanding the search region. To address these two issues, we propose a novel multilevel feature-weighted Siamese region proposal network tracker, which employs a feature fusion module to construct discriminative feature embedding and a similarity-based attention module to suppress the distractors in the search region. Furthermore, a color-based constraint module is presented to further suppress the distractors with the same class to the target. Finally, a well-designed global redetection scheme is built to handle long-term tracking tasks. The proposed tracker achieves state-of-art performance on a series of popular benchmarks, including object tracking benchmark 2013 (0.699 in success score), object tracking benchmark 2015 (0.700 in success score), visual object tracking 2017 (0.470 in expected average overlap score), and visual object tracking (0.485 in expected average overlap score).

The silhouette extraction from surveillance videos is a challenge problem in the field of computer vision. Although there have been many works on this topic, most of them cannot be applied in the practical surveillance scene. In this paper, we propose a hierarchical filling based algorithm to get precious silhouettes of foreground objects quickly. Extensive experimental results show the effectiveness of our scheme.

Sensor planning chiefly applies to optimizing surveillance tasks, such as persistent tracking by designing and utilizing camera placement strategies. Against substituting new optimized camera networks for those still in usage, online sensor planning hereby involves the design of vision algorithms that not only select cameras which yield the best results, but also improve the quality with which surveillance tasks are performed. In previous literatures about coverage problem in sensor planning, targets (e.g., persons) are justly simplified as a 2-D point. However in actual application scene, cameras are always heterogeneous such as fixed with different height and action radii, and the monitored objects has 3-D features (e.g., height). This paper presents a new sensor planning formulation addressing the efficiency enhancement of active visual tracking in camera networks that track and detect people traversing a region. The numerical results show that this online sensor planning approach can improve the active tracking performance of the system.

Frame-skipping videos usually appear in wireless video sensor networks which have wirelessly interconnected devices that are able to ubiquitously retrieve video content from the environment. Frame-skipping videos bring to difficulties in getting the transition model (how objects move between frames). We propose a particle filter with global motion detection requiring no offline or online learning. Experimental results show the proposed approach improves the tracking accuracy in comparison with the existing conventional methods, under the condition of frame skipping data and motion of both targets and video sensors.