Research On Underwater Object Detection And Tracking Method Of AUV

Ocean is the orgin of life.It is an important living space for human with a great of rich natural resources.Because of the particularity in marine environment,it is difficult to complete ocean exploration and development based on human labour.Under these circumstances,many kinds of underwater vehicles appear on the scene.The underwater vision system is adopted for autonomous underwater vehicle(AUV)to sense external conditions.The external environment information provided by vision system is used for planning and decision system of AUV to navigate.Based on the Ministry of Industry and Information Technology(MIIT)project named "ocean exploration research project with autonomous underwater vehicle for engineering",processing methods are studied in this paper for underwater sonar and optical camera images.Besides,object detection with information fusion method and methods of pipeline dectection,recognition,and pipeline location to determine are researched.With the method studied above,a vision system of underwater pipeline detection and tracking is built in this paper.The main research contents in this paper are as following:(1)The characteristics of underwater optical image and sonar image were analyzed.Algorithms of image filtering,enhancement,segmentation and morphological technology are studied aiming at the underwater images.In image enhancement,conventional fuzzy enhancement algorithm is improved to overcome the defects of too many parameters,grayscale losing and low calculation efficiency.The developed enhancement method can keep more characteristic information of object and background.In image segmentation,a segmentation algorithm is presented for underwater images with the feature of low contrast and grayscale overlap of target and background based on two-dimension Abutaleb maximum entropy of image edge gradient information and improved particle swarm optimization(SPSO)algorithm.The segmentation can effectively detect underwater object with low calculation time.(2)Underwater target identification method based on feature fusion is studied.On the basis of sonar and optical image,the Hu invariant moments and wavelet invariant moments of the target are separately extracted.Principal component analysis method(PCA)is introduced for feature dimension reduction in the wavelet invariant moments.Then,based on weight serial feature fusion,an underwater target detection algorithm is established with neural network and SPSO.The recognition experiment results of test samples show that the fusion algorithm has high recognition rate and great generalization ability.(3)Underwater pipeline recognition method of pipeline detecting and tracking system by AUV is researched in this paper.The advantages and disadvantages of the traditional edge detection operators are analyzed.And a kind of pipeline edge detection method is built based on fast Kirsch edge detection operator and Yasuda local edge detection.The search scope of pipeline edge detection is reduced when the discrete kalman filtering model is adopted to establish the correlation gate among the pipeline sequence images.According to the question of bend pipe detection in sequence images,pipe feature area is extracted based on the detection result of Hough transformation.Comparability measure is demonstrated to realize the accurate judgment of the bend pipe in the process of pipeline detection.(4)The coordinate transformatiom between pipeline image coordinates and AUV coordinates is studied to realize underwater pipeline tracking.Considering the camera distortion and calibration accuracy,Zhang Zhengyou’s method is adopted to calculate the parameters of the camera model.The transformation relationship can be determined from the target point of pipeline on the image coordinate system to the AUV coordinate system.And the vehicle can complete pipeline tracking with vision system.(5)Hardware and software architectures of the AUV environment perception system are constructed.Integrate the AUV system and the pipeline detection system.Experiments of AUV pipeline detection and tracking are carried out on the semi-physical simulation platform based on the graphic workstation as well as in the pool environment to verify the validity,realtime performance and reliability of the image processing,detecting and tracking algorithms.