Research On Infrared Sea Multi-target Detection Based On BING Algorithm

Infrared imaginary sea surface multi-target detection technology refers to the given infrared image,the targets to be found,and their locations accurately labeled.Infrared imaging sea target detection technology in the field of computer vision has always been the focus of national defense research in each country,its main application areas are still in the military,but it is also gradually developing toward the civilian direction.The difficulty of infrared sea multi-target detection technology lies in that the size versatility,the size of each target may be different due to the perspective problem and serious influence of sea clutter as well.Traditional methods usually first extract the specific features of the targets,and then train them,get the template that is the feature classifier,and then use this template to test the image window sliding traversal.In this process,there will be many suggestions window,and then use the function of classifier to evaluate these suggestions window,and finally determine the target area by the score to detect the target.In the window sliding traverse process,the number of recommended windows will be more and more,but most of these windows are useless.In order to speed up the target detection and recognition processing,you can reduce the number of candidate windows generated by template traversal.In this paper,we present a multi-target detection method based on BING algorithm for infrared imaging ocean surface based on the current difficulties in the infrared multi-target detection technology and through the reading and analysis of relevant literature.The main contributions of this paper are as follows:In the first part,an infrared imaging ocean surface multi-target detection method based on BING algorithm is presented,which can quickly reduce the candidate target window and quickly detect the target.Combined BING with image saliency detection,a more efficient maritime multi targets detection method is proposed.Compared with the previous method,the results show that the saliency BING strategy can significantly reduce the suggested windows containing objects and thereby be beneficial to the efficiency of detection and targeting.The third part covers an integrated detection and recognition design and experiment in which we adopted the proposed saliency BING strategy to determine the candidate regions,employed the row and column mean based image segmentation to reveal the locations,evaluated circumscribed rectangle to compare with the introduced target feature parameters such as aspect ratio and circularity,et al.