Real-Time Object Tracking Using Dynamic Measurement Matrix
Object tracking has attracted a lot of attention over the past decades. Features represent the main and primary information of object, however, fixed and invariable feature extraction methods would make the features losing their representation. In this pa
- PDF / 1,537,045 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 90 Downloads / 220 Views
Abstract. Object tracking has attracted a lot of attention over the past decades. Features represent the main and primary information of object, however, fixed and invariable feature extraction methods would make the features losing their representation. In this paper, we propose a novel robust single object tracking approach using Dynamic Measurement Matrix to extract dynamic features. In particular, we employ the dynamic measurement matrix to adaptively extract features for discriminating object and background so that features have better and clear representativeness. In additional, our approach is a tracking-by-detection approach via a Naive Bayes Classifier with online updating. Compared to traditional approaches, we not only utilize a Naive Bayes Classifier to classify samples but exploit the nature of this classifier to weigh each compressive feature unit, which would be used to update the measurement matrix. The proposed approach runs in real-time and is robust to pose variation, illumination change and occlusion. Furthermore, both quantitative and qualitative experiments results show that our approach has more stable and superior performance. Keywords: Object tracking · Dynamic measurement matrix Compressive sensing · Dynamic feature
1
·
Introduction
Object tracking is a longstanding problem in computer vision and numerous approaches have been proposed. However, object tracking remains challenging due to the appearance change caused by occlusion and motion and the low resolution. Comaniciu et al. [10] proposed MST (Mean Shift Tracking) approach in tracking, which utilized RGB channels’ histogram to represent and match the target. Briechl et al. [8] used sliding window to search for the target around the object location, which comes from the previous frame. KLT (Kanade-Lucas Tracker) [7] utilized optical flow to find the best area that matches with the affine transformation of the target, and then regarded this area as the object location. Above tracking approaches can track target well in certain situations. However, c Springer Nature Singapore Pte Ltd. 2016 T. Tan et al. (Eds.): CCPR 2016, Part I, CCIS 662, pp. 426–436, 2016. DOI: 10.1007/978-981-10-3002-4 36
Real-Time Object Tracking Using Dynamic Measurement Matrix
427
these approaches still have disadvantages if the object has obvious deformation. These approaches overlook and waste background information, which can improve the efficiency and accuracy of the tracking. Generally, most of these tracking approaches assume the target without obvious and severe deformation during the process of tracking, which can cause tracking drift. Yu Xiang et al. [22] utilized Markov Decision Processing (MDPs) to model the lifetime of an object, and then formulate the problem of online Multi-Object Tracking (MOT) as decision making in MDPs. Xinge You et al. [23] proposed a local metric learning approach to well handle exemplar-based object detection while few exemplars are available. Kiran Kale et al. [19] combined optical flow with motion vector estimation to realize re
Data Loading...
