We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Vehicle-Mounted Infrared Pedestrian Tracking Based on Scale Adaptive Kernel Correlation Filter.
- Authors
Yuanbin Wang; Yujie Li; Qian Han
- Abstract
Pedestrian tracking for vehicle-mounted infrared images is essential in the vehicle-assisted driving system. In general, dealing with the change of the target scale in vehicle-mounted pedestrian tracking is tough. To solve this issue, this study proposes a pedestrian tracking algorithm based on Scale Adaptive Kernel Correlation Filter (SAKCF). The median filtering is applied to suppress the background information and improve the ratio of signal-to-noise. Histogram of Intensity (HOI) feature and Histogram of Oriented Gradient (HOG) feature of the image are extracted and input into SVM. In order to make the tracking adaptive to the scale change of the target, the method of SAKCF is applied. The algorithm is divided into two stages: position detection and scale detection. During position detection, a cyclic matrix around the target position is constructed for dense sampling, with the HOG feature extracted to train the position filter by the output response of Gaussian distribution. During scale detection, multi-scale cyclic shift samples are obtained in the central area of the position detection result and zoomed to the fixed scale of the target by bilinear interpolation, with the value of target regression defined by Gaussian distribution. Experiments indicate that the proposed method can acquire stable tracking on pedestrians for vehicle-mounted infrared images and meet real-time requirements with solid robustness.
- Subjects
PEDESTRIANS; FEATURE extraction; INFRARED imaging; GAUSSIAN distribution; TRACKING algorithms
- Publication
IAENG International Journal of Computer Science, 2022, Vol 49, Issue 2, p349
- ISSN
1819-656X
- Publication type
Article