Abstract:

As the common automatic matching is difficult to meet the requirements of stereo matching for traffic accident investigation, a new automatic matching method of image feature points of accident vehicle is proposed, and the corresponding model is established based on the SIFY theory and the BP neural network. Some experiments are then performed to verify the proposed model. The results show that the proposed method is suitable for the automatic matching of any selected object in a stereopair, and the matching results can be used as a reference to the precise matching. It is also indicated that the proposed method greatly improves the efficiency of stereo keypoint selection and accelerates the measurement in stereovision system. Moreover, as an efficient technology, it provides a reference for the identification of vehicles with covered license.

Key words: traffic safety, stereo matching, SIFT theory, BP neural network, stereovision