A Vehicle Recognition Method Based on Adaptive Segmentation

The rapid development of computer technology makes the realization of intelligent driving system possible, which provides a solution for many traffic problems today. Among them, vehicle recognition is the most crucial. Most of the previous work is based on the computer vision approach. In this work, we propose a novel method of exploring the position information to improve the target segmentation based vehicle recognition. Specifically, besides the vision clues the target position and height are also exploited adaptively in the target segmentation method, so as to get a better segmentation result. Then a multi-class classifier is adopted by using sparse autoencoder based feature extraction. The experiments are conducted on real world dataset. The experimental results show that in contrast with the state-of-the-arts, our method achieves the best performance. The dataset used for segmentation consists of 60 video sequences and the dataset used for identification contains 11173 samples. On these datasets, the proposed method has obvious improvement compared with the traditional method.