Automatic Segmentation Of Aerial Maneuvering Targets Based On Saliency Analysis

As a pre-processing process,image segmentation plays an important role in image processing.Visual saliency target segmentation method is an important branch of image segmentation,which can quickly segment the region or target in the image by simulating human visual attention mechanism.Today,visual saliency target segmentation has been greatly developed,which has been widely used in the fields of target tracking,target recognition and human-computer interaction with the increasing amount of image data.In order to cope with the continuous development of the current military environment and the greatly improved mobility of aerial targets,higher requirements are put forward for the segmentation method of aerial maneuvering targets to achieve faster and more effective segmentation of aerial maneuvering targets.In this paper,a saliency segmentation algorithm for aerial maneuvering targets is proposed in term of the characteristics of aerial maneuvering targets.Firstly,a multi-dimensional mixed Gaussian model with multi-color information fusion is proposed to model the multi-color information of each region of aerial maneuvering target image,which describes the region color information of aerial images more accurately.On this basis,the salient features of the aerial image area are analyzed locally and globally,and the visual contrast between the regions of the aerial image is calculated by using the aggregated WASSERSTEIN distance in the analysis process,which solves the problem that the hybrid Gaussian model can not directly use the WASSERSTEIN distance for high speed calculation.The local and global features of the image are fused through the background prior knowledge to ensure the integrity and accuracy of the maneuvering target area in the obtained salient map.At the same time,in order to further improve the segmentation accuracy of the maneuvering target,the concept of cooperative discriminant factor is introduced to distinguish the maneuvering target during the maneuvering target segmentation process,and the influence of sky background interference on the maneuvering target segmentation accuracy is excluded.In order to ensure the efficiency of the algorithm to the sequence image,the significance analysis area of the maneuver target is updated in term of the significant region position relation between the images in real time.Only the significance analysis of the specified region can effectively reduce the influence of the sky background area on the segmentation accuracy of the maneuver target.At the same time,it also prevents the waste of resources caused by the significance analysis of the background region,and meets the realtime requirement of the algorithm in processing the aerial image sequence.We conducted tests on multiple types of air maneuvering target data sets to prove the effectiveness of the algorithm in this paper.The experimental results show that the algorithm proposed achieves more than 95% accuracy for aerial maneuver target segmentation.At the same time,the aerial maneuver target sequence images with a size of 510 ?300 pixels can be processed more than 30 frames per second to meet the actual project requirements.