Abstract: A novel vehicle detection method based on Haar-like features and Adaboost algorithm is proposed to improve the capability of front vehicle detection of the driver assistance system. First, Haar-like features are selected from the training samples. Then, a learning algorithm based on Adaboost selects the efficient features from the Haar-like feature sets to yield vehicle detection classifier. The classifier is used to examine the testing samples by the pictures provided by the IEEE International Workshop on Performance Evaluation of Tracking and Surveillance. Result show that the proposed method can detect vehicles rapidly and effectively in daytime.

Key words: traffic and transportation safety engineering, vehicle recognition, Haar-like characteristic, Adaboost algorithm