Research on A-pillar Display Algorithm Based on Line-of-Sight Estimation

Abstract:
This paper studied a real-time A-pillar display algorithm based on an improved line-of-sight estimation method. In order to solve the problem of A-pillar blind area, a method based on BoTNet network for line-of-sight estimation is proposed. Compared with the published methods, this method can reduce the average angle error by 9.1%. At the same time, we use two monocular cameras, camera 1 to estimate the driver's line of sight, obtain the coordinates of the human eye according to the line of sight estimation, and use the method of 3D reconstruction of the blind spot to capture the content obtained by the camera 2 in real time and output it to the display screen to realize the blind spot. real-time display. The experimental results show that the frame rate displayed in the blind area can reach the highest frame rate of the network camera of 30fps.