Spatial Analysis Based On Forward Motion Matching

Spatial analysis in complex traffic scenarios,including road vanishing point detection,obstacle detection and spatial depth estimation,is an important part of autonomous driving technology.The existing algorithms of road vanishing point detection,obstacle detection and depth estimation make full use of the static information of the image,and the research of using the motion vector generated in the motion process to carry out spatial analysis is a little insufficient.In this paper,the distribution law of motion vector in forward motion state is analyzed.According to the distribution characteristics of motion vector in forward motion process,a spatial analysis method based on forward motion vector distribution is proposed.The main work of this paper is as follows:1)Aiming at the irregular deformation of the scale and structure in the forward motion,the ORB feature matching algorithm was improved.A progressive scale ORB feature matching algorithm was proposed,and the motion vectors in the motion were obtained from the motion matching results.2)The distribution law of motion vector in forward motion state is analyzed,and the distribution law of motion vector in different planes is shown as large near and small far away.The motion vector in the same plane is linear distribution,and the dynamic vanishing point is at the center point of the image.3)According to the result of motion vector,the dynamic vanishing point is obtained,and the dynamic vanishing point is corrected to the center point of the image to make it conform to the distribution law of motion vector in the forward motion state.It is assumed that the pixels with the same distance from the center point of the image are in the same plane in space.According to this basic assumption,the coplanar model is constructed and the spatial depth estimation algorithm based on the coplanar model is proposed.4)According to the spatial depth value of feature points,the candidate regions of road vanishing points were determined,and the image was partitioned according to the Angle with the candidate points as the center.A road vanishing point detection algorithm was proposed based on the linear fitting of the spatial depth value of feature points.A road area is established according to the vanishing point of the road.A obstacle detection algorithm in the road area is proposed.