Automatic recognition of vehicle image based on monocular vision and environmental perception

Aiming at the problems of low recognition accuracy and long time-consuming in current automobile recognition research, an automobile image recognition method based on monocular vision and environmental perception is proposed. A hybrid filter is composed of median filter and mean filter to suppress image noise and preserve the edge features of the signal. The non-target background is removed by environmental perception, and the target area is obtained with the geometric information in the vehicle shadow as the constraint condition. According to the result of image processing and the determination of target area, HAAR-like feature vectors of targets are extracted and dimensionality reduction is processed. The training classifier is constructed by using the obtained eigenvectors to recognise the current frame vehicles. The experimental results show that the method has the advantages of high recognition accuracy and short time-consuming.