Visual information processing using cellular neural networks for mobile robot

Abstract

Visual information processing is one of the key technologies for robot visual navigation, whose speed directly determines the quality of the visual navigation. Taking advantage of the parallel image processing capability of cellular neural networks (CNN), we propose a fast algorithm using CNN for mobile visual information processing. In the algorithm, convex restoration, gray threshold, dilation and erosion, and edge detection using CNN are performed to achieve road image filtering, image segmentation, edge detection, and other image processing operations respectively. Experimental results demonstrated that the CNN has strong image processing adaptability, which can fast achieve structured and unstructured roads filtering, image segmentation, and edge detection. The proposed method can eliminate the influence of shadows and water marks on the segmentation of road images, and can segment and detect the lane area quickly, effectively and robustly.