Research On Ship Target Detection And Recognition In Sea Battlefield

As one of the main combat zone in modern warfare,the sea battlefield situation is varying from minute to minute.Ship target is the key target of maritime monitoring and wartime combat.And the outcome of battle is greatly determined by whether to identify ship target quickly and provide support for commander in decision-making or not.The monitoring system determines the ship's tactical intent which is based on accurate detection and recognition of ship targets.Only accurately detecting and identifying the ship target battlefield attended,the monitoring and analysis system can make an accurate analysis and prediction of ship targets tactical intentions and assist commanders to make the right decisions.According to the needs of the ship targets detection and recognition of the sea battlefield in naval Command & Control,this article studies the problem of ship targets detection and recognition in high resolution optical remote sensing images.The main innovation points of the article can be described into three aspects as follow:(1)A ship target detection method based on visual saliency is proposed,which is using the ship target characteristics in the optical remote sensing image to establish the ship targets saliency detection model.Firstly,the model extracts image saliency information by means of two feature extraction methods which are based on bio-inspired and pure mathematical calculation,the former mainly through the significance of local regions calculates the significant information of image,the latter calculates the significance of the image information via the global information of the image.Then,it creates visual saliency feature selector,introduced the concept of cluster density,with the cluster density of different feature maps to describe the significant of each map and calculate the feature weights,and to choose the higher weights five feature maps.Finally,the five weighted maps are combined into a visual saliency image.Targets will be extracted from the saliency image.The advantage of this method is that the selected feature channel is suitable for extracting ship targets.Significant target image can be more accurately described by the above two feature extraction methods.The use of visual saliency feature selector can choose the most effective feature maps and combine them into a saliency map.(2)The target edges to be identified of the traditional methods extracted from the image are often not smooth;there may be holes in the object contour area to be identified.These factors are not conducive to the target feature extraction,and ultimately may affect the accuracy of target identification.In response to the problems,this paper presents an imagecontours simplified method which is based on Hough transform neighborhood.Different from the traditional Hough Transform applications,the method makes use of the basic principle of Hough transform,introducing the concept of neighborhood in the parameter space.The straight line in the image space is mapped to a point in the parameter space,in this point as the center to establish a sufficiently small neighborhood.When the all curves in the parameter space which are mapped by the points in the image space are through the neighborhood of parameter space,it can be approximately considered that the points in the image space are in the same straight line,thereby determining the set of points in the image space that can be approximated as a straight line.According to the maximum distance method taking two points as the starting point and end point of a line and these points sequentially connected to form a closed curve composed of the segments,so we can get a simplified image contour.The advantage of this method is that you can easily simplify image contours,and make simplified contour approximation of the original full image contour.(3)In modern naval warfare,ship targets are the sea battlefield focus.Ship targets in different types need to use different way to attack.In order that target strike mission is accurately completed,the key problem is to identify the target.In response to these problems,a ship target multi-feature fusion recognition method based on DSm T(Dezert-Smarandache Theory)is proposed in this paper.The method comprises three main parts: a)It establishes the DSmT fusion recognition model,that uses geometric features and shape features of ship targets as evidence sources.These features are simple,easy to access and in ship target recognition with fine discrimination.b)It defines PNNs(Probabilistic Neural Networks)for ship features,through neural network training,BBA of each feature can be got.c)According to the characteristics of the ship target recognition model,it analyzes the robustness of different PCR(Proportional Conflict Redistribution)rules and uses a more robust to fuse the BBA of each evidence.In recognition of ship targets,this method has high recognition accuracy.