Self-adaptive cluster segmentation aircraft objects in aerial images
Yingchun Li, Hexin Chen, Ming Zhao, Pengfei Qu · 2004
We present a self-adaptive cluster segmentation method for the problem of automatically detecting the aircraft location from complex aerial images. Image segmentation that partitions a given image into meaningful regions is an important task of image analysis for recognition. We introduce knowledge about the location of the object of interest and knowledge about the behavior of edges in scale space, in order to enhance edge information. We present a new region potential term based on the classical iterative method when it is applied to the segmentation of aerial images. Following an edge-linking procedure, the regions of objects can be bounded by closed boundaries. The method is applied to a number of aerial images, each one of which contains one or more objects. Experimental results are provided to illustrate the correction of this object detection method in a lot of domain, regardless of the complexity of background in images.