Indoor Robot Visual Positioning System Based on Floor Features

In order to solve the positioning problem of indoor robots, we present visual positioning system for indoor robots based on ground features. The system firstly collects the ground image through the fisheye camera, then uses the Canny operator to perform edge detection on the captured images, then performs the distortion correction of the fisheye camera on the contour points obtained by the edge detection. It restores the contour points to the undistorted plane. Using RANSAC algorithm and least squares method to compute the relative position of the camera to the ground line segment feature contained in the contour, then using the IPM algorithm to transform the feature into the real world space. We finally obtain the distance and orientation angle from the camera to the ground feature segment. Since we measured and stored all the feature features into a map beforehand, this method can locate robot position when a feature segment is observed. Experimental results show that the positioning method has high positioning accuracy, good real-time performance even with embedded CPU and low cost, so it can solve the positioning problem of indoor robots well.