Road extraction from high resolution remote sensing image using multiresolution in case of major disaster

Abstract

Road extraction is a topical research because of complexity due to his large topological variability. Increasing the spatial resolution generates noise which makes extraction difficult, especially in case of major disaster in an urban context. This problem increases false alarm rates and generally affects the performance of road extraction algorithm. Our aim is to improve the quality of roads extraction after adaptation of the Lowe's SIFT descriptors (scale-invariant feature transform) jointly with spectral angle algorithm. The characterization is performed on two image at various resolution images, respectively representing a rural and urban disaster area, captured by Quickbird satellite. Our approach significantly reduces the amount of false detection and shows an overall accuracy of up to nearly 30% in some cases.

Road Extraction From Very High Resolution Remote Sensing Optical Images Based on Texture Analysis and Beamlet Transform

“…In terms of correctness, our proposed algorithm achieves better results and does not produce false alarms, in contrast to Coulibaly et al's method [48], which generates several false alarms, especially for study area 2. In terms of redundancy, both algorithms obtain very good results for all three study areas.…”

“…In terms of redundancy, both algorithms obtain very good results for all three study areas. As with completeness, the proposed algorithm outperforms Coulibaly et al's [48] method for study area 1 and study area 3.…”

“…The proposed algorithm returns better results for study area 1 and study area 3. For study area 2, Coulibaly et al [48] obtain a higher value for completeness in a rural area. This is attributed to the small width of the road surface in the image and the high contrast of the road, conditions which favor Coulibaly et al's [48] method.…”

“…In order to evaluate performance of the proposed road extraction algorithm, we implemented the method proposed by Coulibaly et al [48] based on the spectral angle algorithm and multiscale analysis. Extracted subimages of an urban area [ Fig.…”

“…Comparison between the results of the proposed method and Coulibaly et al's[48] method: (a)-(c) represent study area 1: urban region, study area 2: rural region, and study area 3: peri-urban region, respectively, and (d-f) their associated road vectors. The results of Coulibaly et al's method[48] for the three areas are shown in (g)-(i), and (j)-(l) indicate the results of the proposed method for the same three study areas.…”