Man-made object detection in aerial images using multi-stage level set evolution
The detection of man-made objects is important for scene understanding, image retrieval and surveillance. A modified Chan-Vese model based level set method for detecting man-made objects in aerial images is presented. The method applies a fractal error metric with an additional constraint-texture edge descriptor to realize a more accurate segmentation. Man-made objects are extracted from the natural areas by changing the geometric active contours, which are governed by a partial differential equation based on the modified Chan-Vese model. Benefiting from the level set method, the extracted man-made object contours may change topology easily during the evolution. Our method does not need to select a threshold to separate the fractal error image in cases where segmentation errors may result from using an unsuitable threshold. We validate the proposed algorithm by numerical results of real aerial images.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
No Article Media
Document Type: Research Article
Affiliations: Institute of Image Processing and Pattern Recognition, Shanghai Jiao Tong University, Shanghai 200240, PR China
Publication date: January 1, 2007