Research On Target Recognition Method Based On Shape Appearance For Unmanned Ship

Along with the rapid development and innovation of new technologies and new concepts such as computer technology and artificial intelligence,autonomous navigation has become one of the important directions for the development of ships.In order to be able to achieve autopilot or perform some special tasks,the unmanned surface vessel(USV)uses advanced testing equipments and methods,completes the required actions according to different tasks,ultimately realizes the autonomous navigation of the ship.In the course of autonomous navigation,unmanned ships need to use their own detection equipment to perceive and judge the external environment.Among them,it is one of the most important tasks for the unmanned ship to recognize the target.Based on the above background,this paper makes use of the water surface imagery collected by unmanned ships to study the target recognition algorithm of marine vessels.This paper mainly includes the following aspects,its content is summarized as follows:First of all,this paper carries on the characteristic analysis to the water surface image,and according to the characteristics of the surface image,the preprocessing method of water surface image is divided into two kinds: the image filtering algorithm and the image enhancement algorithm.In this paper,the adaptive median filter algorithm is used to denoise the image,and water surface image is enhanced by adaptive platform histogram equalization algorithm.The pre-processing operation of the surface image can greatly improve the quality of the image,and prepare for the subsequent ship recognition.Secondly,in the surface images collected by the unmanned ship,sea antennas often appear in its background images.In view of this characteristic,the detection algorithm of sea antennas in the water surface images is studied in this paper.Through the simulation comparison test,it can be seen that the use of Hough transform to detect the sea antenna can get better detection results.The detection of sea antennas can not only reduce the influence of sea wave clutter and other interference information on target recognition,but also accelerate the speed of target recognition and improve the recognition efficiency.Thirdly,in this paper,the image segmentation algorithm based on genetic neural network is applied to the segmentation of water surface image.The target ship can be separated from the background image directly,and it is not disturbed by the noise of the sea wave and other environmental noise.Simulation results show that the segmentation algorithm based on genetic neural network is better than the one based on BP neural network to segment the watersurface image and the traditional method to segment the water surface image.Then,using the geometrical features and the HU invariant moments to extract the features of the ship images,the paper extracts and simulates the features of the ship images.From the result,it can be seen that using the geometrical features and the HU invariant features can well realize the ship image characterization and classification.Finally,in this paper,Support Vector Machine(SVM)is used to train sample data in ship characteristic information base,and cross validation is used to optimize parameters in support vector machine.Finally,the MATLAB simulation program is applied to classify and identify different types of ship images.