Visual Odometry.ppt

Result: SIFT matches SIFT features matched and inliers identified using RANSAC Improvements Gaussian convolution of produced data-points Self-tuning of RANSAC parameters More intelligent sliding window for bundle adjustment Hidden Slide Chris 40% Mark 30% Firouzeh 30% Visual Odometry Chris Moore Mark Huetsch Firouzeh Jalilian Problem: Estimate camera motion Estimate trajectory of a vehicle which captured sequence of images from all four sides Approach Compute SIFT features for a stream of frames from a camera Set a window size for the number of consecutive images to handle at a time For each window of n frames For first 2 frames Find corresponding SIFT features (rough correspondences) Calculate essential matrix using RANSAC to discard inaccurate correspondences Decompose essential matrix into rotation and translation, and calculate sparse 3D structure (Can only find translation and structure calculate up to a scale!) For later frames i=3…n Find rough SIFT correspondences between frame i and frame 1 Calculate motion from frame 1 to frame i using 3D structure computed in a Use sparse bundle adjustment on all frames in window to refine camera motions and calculate scale Advance the window by 1 frame, repeat step 3