Environmental knowledge-guided bridge detection over dry-season-river from high resolution remote sensing image

A detection method for bridge over dry-season-river is proposed in this paper, from high resolution remote sensing (RS) image. Traditional bridge detection methods excessively depend on characteristics of water. However, when the river is in the dry season, water characteristics will not be reliable, and large errors will be obtained on the riverside-part. To solve this problem, a two-stage detection system is introduced. In first stage, water characteristic based method is reused for initial locating the position of bridge, then a shadow characteristic constrained criterion (SCCC) is proposed for accurate correcting, which utilize shadow of bridge to optimize the detection. Experiments illustrate that comparing with traditional method detection accuracy have obviously improvement with our method.