Ship image object detection based on the fusion of visual saliency and sparse learning
In order to suppress the influence of light change,wave interference,background clutter and other factors in ship image target detection,a ship image target detection method combining visual salience and sparse representation learn-ing is studied to improve the ship image target detection effect.Using ship image to build ship image dictionary;The ship image is sparsely encoded by sparse representation algorithm combined with dictionary.Based on sparse coding results,visu-al significance maps are extracted from ship images.Through the adaptive threshold method,the visual significance map is segmented to obtain the candidate region of ship target and narrow the detection range of ship target.In the probabilistic neural network,the candidate region of the ship target is input to judge whether it is the ship target,and the ship image target detection is completed.Experimental results show that the proposed method can effectively sparsely encode ship images and extract visual significance images.The method can effectively segment the visual significance map.This method can accur-ately detect ship targets in both simple and complex backgrounds.
NETL
NSTL
万方数据
