Method for the Extraction of Loess Shoulder-Line from Grid Dems Based on Log Edge Detector

This paper proposes a new method for extracting loess shoulder-lines from grid DEMs. The morphological characteristics of loess shoulder-lines are investigated firstly. Loess shoulder-lines is the border line of loess inner gully area and inter gully area and its upper side is the relatively flat area(with slope less than 20 degrees) called Inner gully, the lower side is the inter gully with slope between 30 degrees and 35 degrees. There is remarkable morphological variation between the upper side and the lower side. The variation is a clue to extract candidate points of loess shoulder-lines. By applying the edge detection operator LOG to noise removed DEM, the method for extracting loess shoulder-lines candidate points is proposed based on the prominent height variation of the points along the shoulder-lines. This method can detect shoulder-lines described above. The algorithm then connects the candidate points to small line segments by morphological method which labels points in an eight neighborhood window as a line, and finally, precise and systematic loess shoulder lines are extracted after refining the line segments by a simple filter of line length. Experiments in the loess hilly areas show that the loess shoulder-lines extracted by LOG operator has very good matching ratio to standard shoulder-lines extracted manually and good performances on line integration. This ratio is achieved by the so called contour matching method for line matching. Our experiments show that the LOG edge detector has the potential of detecting the explicit shoulder-lines by the zero cross of height second derivatives and the morphological filter can restrain the affection of random noise. The matching scores of extracted shoulder-lines with the standard shoulder-lines from real terrain show the LOG edge detector is a good shoulder line extraction method.