An effective algorithm for tank objects identification in a complex background that fea-tures the perceptual organization is proposed in this paper.With multi-window architecture,the algorithm consists of two parts:coarse recognition and detailed recognition.Based onprior knowledge,coarse recognition scans the entire image data,then gets the target-kerneland its interesting window to direct detailed processing.Moreover,the detailed recognitionexecutes a depth-first search which retrieves locally around the target-kernel in the windowaccording to the rule of similarity measure.Experimental results show that the algorithm canidentify tank objects in a complicated scene effectively.
展开▼