Research on Blind path Obstacle ranging based on improved ORB Feature Matching

Abstract

For the problem of poor real-time and accuracy of distance measurement of obstacles on blind corridors with the traditional ORB (Oriented FAST (Features from Accelerated Segment Test) and Rotated BRIEF (Binary Robust Independent Elementary Features)) matching algorithm, a method of narrowing the search domain is proposed to optimize the ORB algorithm. Firstly, the image is fed into the trained SegNet network model to obtain the segmented blind image, which is converted into a binary image to obtain the position information of the blind in the x-direction of the whole image, and then feature point extraction is performed in the blind area. Secondly, the polar line constraint is added to the feature matching part and the RANSAC algorithm is combined to optimize the matching to obtain the obstacle feature point pairs in the blind area. Finally, the parallax is calculated according to the two-dimensional planar structure of binocular stereo vision to find the precise distance of the obstacle. The experimental results show that the improved ORB algorithm has better real-time performance and accuracy.