A robust autonomous mobile forklift pallet recognition

Abstract

A visual pallet recognition system based on the accurate color segmentation is proposed. The basic idea is to get the pallet color feature samples from the images in working environment before the autonomous mobile forklift works. Then the algorithms of morphological filtering, Sobel edge detection and Hough tranform are used to get the pallet forking side. At last, according to two corners of the forking side, the pallet midpoint coordinates of the forking part and the direction of the forking side are calculated, which are the pose provided for the autonomous mobile forklift engagement. Experimental results show that the visual recogition system has good real-time and robustness.