JCSE, vol. 12, no. 3, pp.127-138, September, 2018

DOI: http://dx.doi.org/10.5626/JCSE.2018.12.3.127

Vision-Based Blind Spot Monitoring Using Rear-View Camera and Its Real-Time Implementation in an Embedded System

Kyeong Hoon Jung and Kang Yi
School of Electrical Engineering, Kookmin University, Seoul, Korea School of Computer Science and Electrical Engineering, Handong Global University, Pohang, Korea

Abstract: Blind spot monitoring (BSM) is one of the essential functions of Advanced Driver Assistance Systems (ADAS). In this paper, we propose a vision-based monitoring algorithm to detect vehicles in a blind-spot area using a rear-view camera. Instead of moving around the whole image to search for a vehicle, we use well-defined detection windows with fixed position and size. And we used histogram of oriented gradients (HOG) and the support vector machine (SVM) to detect a vehicle in each detection window. When a vehicle is detected, we compute the motion vector of the vehicle to determine if it approaches the ego vehicle or not. The alarm signal is finally generated based on history of events. To evaluate performance of the BSM algorithm, we captured various kinds of sequences by using a rear-view camera. Experimental results reveal precision higher than 98% and recall higher than 99%. We also develop high-speed methods to improve processing speed while keeping performance degradation as minimal as possible. Implementation of this fast algorithm on a commercial embedded device validated real-time characteristics of the proposed BSM algorithm.

Keyword: Advanced driver assistance system; Blind spot monitoring; Vehicle detection; Intelligent vehicle

Full Paper:   352 Downloads, 1197 View