A Method of Linear Feature Extraction from High Resolution Aerial Images

In this paper, we propose a novel method of gray-level projection called circular projection that is invariant to translations, scale changes and rotations. In particular, the proposed method can reduce the problem of 2-D patterns into that of 1-D ones. The method is applied to feature extraction for road and river from high resolution aerial images. Wavelet transform is then used to analyze the obtained feature curves. It´s observed by experiments that the method combined with wavelet processing can provide reliable features for segmentation and recognition of roads and rivers.