Proposal of low-resolution vehicle image recognition method

Developing on-board automotive driver assistance systems aiming to alert drivers about driving environments and possible collision with other vehicles has attracted a lot of attention recently. Especially, many researchers have proposed the forward vehicle recognition method by a camera on vehicle. In the forward vehicle recognition, however, it is difficult to detect the features of vehicle from a distant vehicle image by conventional methods because the image is too low-resolution (LR). This paper presents a vehicle image recognition method for detecting features of a distant vehicle by frame-merging of image sequences. To detect vehicle features of a distant LR vehicle image, we use the image sequences obtained from the camera on the vehicle, and utilize super-resolution (SR) image reconstruction. In SR image reconstruction, signal processing methods are utilized to obtain a high-resolution image from observed multiple LR images. Through the experiments using real road images, we show the effectiveness of the proposed methods.