Rényi entropy-based saliency map generation and target detection

For improving the accuracy of automatic target detection in complex ground scenes, the statistical property between the Rényi entropy and its occurrence probability of images calculated through the Pseudo Wigner-Ville Distribution (PWVD) with a two dimension window was analyzed based on exponential function. Then the change of the statistical property caused by the appearing of man-made targets is investigated. The methods of saliency map generation and target detection based on Rényi entropy are proposed. Firstly, the image of Rényi entropy is smoothed by the average filter. Then, the image of residual Rényi entropy is obtained by the subtraction of fore-and-aft filter images, and the saliency map is obtained by Gaussian filter. Finally, the target detection is completed by segmenting the saliency map with a simple and convenient threshold method. Experimental results demonstrate that the detection probability and false alarm probability of the method are 100% and less than 7.1% respectively for 14 targets in 8 images. In comparison with traditional methods, the proposed method can detect the military targets from complex ground scenes effectively.