Runway detection using line segment statistical model

To meet the needs of precise runway detection in remote sensing images, while reducing unnecessary adjustable parameters, an efficient and parameter free method is proposed to extract the straight edge of runway based on Helmholtz principle. By introducing the model of line segment model, we compute and analyze the gradient of the pixels on a given oriented line, and then the problem of line segment is classified into hypothesis test under tolerant significance. In theory, this proposed method can limit the number of false alarms and turns out to be parameterless. Using the end point of detected line segment, the extraction of parallel line is of light computational burden. The results of concrete surface and pitch surface airport show our method avoid initializing the type of runway, and is not sensitive to noise and the gray-level difference between airport and ground. The final allocation of runway in an image is accurate and the turning around time is greatly condensed.