OBSTACLE RECOGNITION METHOD FOR LIBRARY ROBOT BASED ON DEPTH IMAGE

摘要

The disclosure discloses a method for detecting an obstacle based on a depth image. The method includes: step 1, choosing the pixels corresponding to a certain distance range from the collected image; step 2, denoising the image and smoothing the edges by using an erosion and expansion operation; step 3, conducting edge detection by a convolution operation; step 4, finding out the bookshelf plane, and drawing a dividing line; step 5, finding out the plane of a protruding book according to a convolution resu and step 6, filtering out the regions with a small area, and judging if there is an obstacle in the regions with large areas.