Adaptive pedestrian tracking via patch-based features and spatial–temporal similarity measurement. (May 2016)

• Record Type: Journal Article
• Authors: Shen, Xuewei; Sui, Xiubao; Pan, Kechen; Tao, Yuanrong
• Is Part Of: Pattern recognition. Volume 53(2016:May)
• Page Start: 163
• Page End: 173
• Publication Date: 2016-05
• Subjects: PH perceptual hashing -- SVM support vector machine -- MIL multiple instance learning -- DCT discrete cosine transform -- HOG histogram of oriented gradient -- SAB Shadow Detection and Removal from a Single Image Using LAB Color Space -- ST Shadow Detection and Removal from a Single Image Using LAB Color Space -- ALSA Visual tracking via adaptive structural local sparse appearance model -- CT Real-time compressive tracking -- OMSIT Online multiple support instance tracking -- IVT Incremental learning for robust visual tracking -- SR success rate -- CLE tracking localization error.; Pedestrian tracking -- Image shadow removal -- Patch-based target representation -- Similarity measurement; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2015.11.017
• Languages: English
• ISSNs: 0031-3203

Abstract

Abstract: The existing pedestrian tracking algorithms are still challenging under abrupt illumination, occlusions, out of field of view, and cluttered backgrounds. In this paper, we propose a novel tracking framework to handle the above tracking failures. The framework consists of two steps: image shadow removal and tracking by association. Shadow removal tackles tracking failures under large illumination variation between shaded and non-shaded area. Tracking by association is based on particle filter. Inspired by the observation that the different parts in one pedestrian target have different stable properties, we represent the target with perceptual hashing (PH) and color self-similarity at patch level, and design an adaptive patch-based similarity measurement which simultaneously considers the spatial and temporal appearance information of the target. Quantitative and qualitative evaluations on several sequences show that our method consistently outperforms tracking methods compared. Highlights: A patch-based appearance model is proposed according to stable properties. The number of valid patches in particles is automatically adjusted. A spatial–temporal similarity measurement is formulated for tracking and updating. A shadow removal algorithm makes tracking robustly under abrupt illumination. An extensive comparison with state-of-art methods.