Haar와 Lab 색상 특징 기반의 야간 차량 전조등 및 후미등 감지 알고리즘

초록

On-road vehicle detection is an important technique in Intelligent Transportation Systems. Under the daylight condition, the vehicle detection system can utilize the appearance information of vehicle such as color, shape, or typical vehicle patterns for vehicle detection with high performance. However, most of vehicle appearance features are insufficient and unstable in nighttime, and headlights or taillights become the reliable features for identifying vehicles. In this paper, a method for detecting both headlight and taillight regions during nighttime is proposed. At first, the Haar-like features using grayscale image and color feature using Lab color space are combined. Then, the Multi-Adaboost is trained with the combined features. The experimental results show the effectiveness of our method in detecting headlights and taillights at night.

키워드