Remote Sensing Image Target Recognition Based On Mathematical Morphology And The Watershed Algorithm

The technology of object recognition in remote sensing image is a very active research field in recent years. The paper combines the mathematical morphology and watershed algorithm to bring out new object recognition method.First, the paper uses the watershed algorithm to segment SPOT remote sensing image. Second, using threshold segmentation in every segmented regions. During the realization progress, the original image has some noise and slim variations in gray value, so it maybe has many false regional maxima. If using watershed transform to the pre-progressed image directly, the image can be over-segmented. For this reason, the paper adopts the morphologic gray reconstruction to remove the regional maxima of the image, in the end, recognize the object based the feature about size, shape and relationship in space.The method that the paper brings out can be used more generally than traditional reorganization method. The method can be used in SPOT RM image with different gray feature, so the method can avoid complex threshold selection progress and make the object recognition automatization more easily.