Extraction building based on adaptive detection algorithm of key points

This paper presents a method that apply shape context features to identify building from high remote sensing image automatically. This method use adaptive algorithm of key point detection to detect the key point of building boundary rapidly. By eliminating the unstable points, the stability of detected edge points can be improved. We apply the selected points for log-polar coordinate´s transformation and the target shape context features calculation to calculate the shape histogram. Lastly, identify the similarity between the shape histogram and the shape context features library made beforehand. The experimental result show that key points detection algorithm have a fast speed, the similarity calculate by the shape context features based on selected boundary points reflect the categories of building and make the buildings have strong separability. Secondly, the shape information of building is available; we have no need to use other software to get the shape information of building.