Aiming at the problems of long-time, high-energy consumption and missing information for labyrinth robot to obtain labyrinth information independently, this paper proposes to use image processing related methods to obtain labyrinth cell information, so as to improve the labyrinth task efficiency of labyrinth robot. Aiming at the inherent defects of uneven brightness and color dissimilation in labyrinth images, a labyrinth cell edge detection method based on Kirsch and improved Canny operator is proposed, which realizes the edge feature extraction of labyrinth cells. Using the principle of image morphology, the labyrinth cell features are denoised, merged, expanded and filled to achieve image enhancement of the labyrinth cell. Aiming at the characteristics of labyrinth cells, a pixel block matching feature method is proposed to match the information of labyrinth cells to achieve the information extraction of labyrinth cells.The experimental results show that the accuracy of information extraction for labyrinth images with a brightness range of γ =1. 1 and a shooting angle of 30 degrees or more reaches 100%. Therefore, the labyrinth information extraction method proposed in this paper is practical and feasible, which provides a guarantee for the efficient execution of the subsequent labyrinth tasks of the labyrinth robot.