Research on unstructured road detection algorithm based on improved morphological operations

Abstract

A new method with improved morphological operations on unstructured road is introduced in this paper. Mathematical morphology erosion and dilation operations are usually used so that the road and background are separated more clearly to acquire obvious road boundaries. How to use these operations reasonable will directly affect the subsequent detection. Firstly a binary image from an original image is segmented into the road and non-road regions by using 2-dimentional Otsu adaptive threshold segmentation algorithm. In terms of the feature of unstructured road, erosion operation is used twice and dilation operation is used once in this paper. Then, LOG operator is applied to detect edge. Finally, Hough transform is adopted to detect and mark the road boundaries. Experiments indicate that our method not only removes the unfavorable elements of non-road areas, but also has the advantages of fast operation and high accuracy.