Video Image Stabilization Algorithm for Remote Control Driving

Under the condition of remote control driving, the vehicle jitter caused by bumpy road condition causes the camera to acquire the picture to shake violently. In order to solve this problem, this paper adopts the method of combining vehicle-mounted inertial navigation and electronic image stabilization to stabilize the image and then compensate the image, so as to realize the stable and reliable display of the image. Firstly, vehicle inertial derivative data and video image data are synchronized in time to determine the base frame. Then Shi-Tomasi corner detection is carried out on the video image, and then the corner is tracked by optical flow method. The affine transformation matrix between adjacent frames is estimated by RANSAC algorithm, and the original affine transformation parameters are obtained. After smoothing the original affine parameters by Kalman filter, the optimized affine parameters are obtained. Finally, synchronous inertial derivative data is used for motion compensation to eliminate the black edge of image stabilization, so as to obtain a stable video, which is further output through the ROS system. The experimental results show that, after image stabilization, the MSE decreases by 996.499 and PSNR increases by 7. 199dB compared with the original video, and the average processing time of a single frame is 21. 855ms.