Road extraction using modified dark channel prior and neighborhood FCM in foggy aerial images

Abstract

It is still a challenge to accomplish road extraction from the aerial images in foggy weather. In this paper, a road extraction method based on modified dark channel prior and improved neighborhood FCM (Fuzzy C-means) is proposed for foggy aerial images. Firstly, a defogging method based on modified dark channel prior is applied to increase the image contrast and highlight the road areas. In the proposed method, a region filtering function is designed to generate the dark channel image, and an adaptive parameter is set up to adjust the defogging degree automatically. Secondly, an improved neighborhood FCM algorithm is studied to extract the road area. According to the neighborhood feature, the spatial distance and the gray level difference can be calculated as the parameters of the objective function which can eliminate noise and promote the detection accuracy. Finally, an image post-processing procedure is utilized to connect the road gaps and remove the false road areas. The experimental results verify that the proposed method can achieve satisfied road extraction effect both on completeness and correctness.