Stereo visual odometry method based on image gradient joint optimization

摘要

A stereo visual odometry method based on image gradient joint optimization includes steps of: pre-processing each frame of images of a stereo video input; calculating a stereo image disparity map of a stereo video frame, and obtaining a metric distance; scaling to form scale images with different scales for forming a pyramid model, and calculating features to obtain a series of feature images; processing the images with a Kalman filter according to the metric distance, so to predict and estimate a motion process of a camera pose, wherein camera motion models is built in the Kalman filter; accurately calculating a camera pose of a current frame by using a gradient-based stereo visual navigation method; and updating the camera motion models in the Kalman filter with the camera pose of the current frame.