Scale-adaptive Tracking Based on Perceptual Hash and Correlation Filter

Wei Huang,Lingpeng Lin,Tianqiang Huang,Jing Lin,Xueli Zhang

DOI: https://doi.org/10.1007/s11042-018-6956-7

IF: 2.577

2018-01-01

Multimedia Tools and Applications

Abstract:In the research on computer vision, object tracking has encountered various challenges, such as occlusion and scale variation. In recent years, tracking-by-detection methods have performed competitively. Some of these methods have focused on solving the problem of scale variation. Regardless, these algorithms perform poorly in real time. Recently, correlation filters have been widely used in object tracking because of their high efficiency; however, conventional correlation filter-based trackers cannot handle scale variation. Most correlation filter-based trackers update the template for each frame, and tracking offsets occur when a tracking error is present. To overcome these problems, we propose a novel scale-adaptive tracking algorithm that uses perceptual hash and correlation filter on the basis of tracking-by-detection methods. We employ kernel ridge regression to minimize the mean square error between the training image and the regression object, and construct a robust filter template to track the target center location. By tracking the 4 sub-blocks of the target image, the length and width expansion coefficients are calculated separately to update the target scale. We finally use the adaptive update strategy based on perceptual hash to effectively prevent the tracking offset caused by the template update error. Owing to the insensitivity to the scale variation and high efficiency of the perceptual hash, tracking becomes more robust in real time. Both quantitative and qualitative evaluations on Object Tracking Benchmark (OTB) indicate that the proposed tracking method performs more favorably compared with other state-of-the-art methods.