Mobile Camera based Motion Segmentation by Image Resizing

Authors
Chunyu Yu, Fengzhi Dai
Corresponding Author
Fengzhi Dai
Available Online 1 September 2016.
DOI
https://doi.org/10.2991/jrnal.2016.3.2.7How to use a DOI?
Keywords
Mobile cameras, motion segmentation, edge-based alignment, object detection.
Abstract
The need of detecting moving object like human and vehicle by mobile camera is increasing in commerce and industry. Therefore, tradition method of abstracting moving object from still background cannot solve the problem. Though Histogram of Oriented Gradients (HOG) have been widely used for human detection, due to relying on sliding windows and multi-scale resizing method, it is computational expensive to perform on ordinary equipment. In this paper, an image-resize methodology which can abstract motion segmentation and detect moving object from moving background is proposed. First, edges images are computed. Then movement vector between frame images are computed and the relative background motion is compensated. By adjusting the parameters of resize algorithm, human liked object or vehicle liked object can be segmented separately and the segmentation can be used for further detection. Experiments have been performed under three different environments for human detection and vehicle detection. The results show that the running time is highly reduced and the accuracy can reach as high as 93.04%.

Copyright
© 2013, the Authors. Published by ALife Robotics Corp. Ltd.
Open Access
This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).