Small target recognition method on weak features

For real time tracking the moving small target under complex backgrounds, this paper will present a tracking algorithm and a recognition method for small target based on the fusion filtering of single frame and multiple frames. For a case that the moving small target is masked or submersed easily by other objects or noise in complex background, this paper will present an opening and closing transform algorithm to eliminate or weaken background and noise; present a competitive model with adaptive neural network of online learning for the puniness characteristics of moving small target, use its competitive active unit to extract the multidimensional characteristic parameter of small target; consequently, present a tracking recognition method for moving small target based on a complex background. In conclusion, the significant novelty is the small target recognition method proposed, including the eliminating noise method, extraction method for weak small features, establishment of prediction model of motion target state, recognition method for small target. These researches development in this paper help to the personnel of target recognition and image processing understand the motion law, reduce expenditure, decrease unnecessary damage or injury, etc.