DETECTION OF THE FRONT VEHICLE FROM THE STEREOSCOPIC IMAGE USING HIERARCHY PROCESS
Nobuhiro Tsunashima, Masato Nakajima · Machine Vision Applications · 1998
In this paper we described a new technique for measuring a distance to the front vehicle using stereo images. Only vehicles traveling in the driver's lane need to be detected. Thus the driver's lane is determined as the processing region in the images. Usually detecting lane markers does this. However, when the front vehicle appears in the near sight, it is not necessary to detect lane markers to the far sight. On the other side, when the front vehicle appears in the far sight, it is necessary to detect lane markers to the far sight. So we adopt the hierarchy process to detect the front vehicle. 1 Introduction prevent the collision accident it is necessary always to calculate the distance from the driver's vehicle to the one in front. So stereoscopic image processing has been used to calculate this distance. Only vehicles traveling in the driver's lane need to be detected. Thus the driver's lane is determined as the processing region in the images. Usually detecting lane markers does this. However using those makers causes the following problems. On a curve, most of the front of the driver's vehicle is outside the processing area. As a result, vehicle can't be detected. When the processing area is expanded outside the lane to adapt for curves, side walls of the road may be detected as the vehicle. Using stereoscopic image processing, it is very difficult to detect only the vehicle in front of the driver. This paper proposes the new method for detecting a vehicle in front. First the disparity map is calculated from the stereoscopic images.