Research On The Method Of Navigation And Positioning For AUV Based On Feature Extraction

In recent years, with the development of autonomous underwater vehicle(AUV) widelyused in the field of marine technology such as military and civilian application. Theautonomous navigation and its error correction ability is one of the key technologies for longtime underwater task. Because the error accumulation of traditional inertial navigationincreases with time during the task, AUV needs to periodically move to surface to use othermethods(such as GPS) to complete the correction of error. It is not conducive for AUV tooperate for long time and affects the effect. Simultaneous localization and mapping method isproposed to make up for the lack of inertial navigation methods. Using information obtainedby onboard sensor to build incremental map, SLAM algorithm does not require a priori map.This paper studies on the navigation of AUV based on extended kalman filter (EKF), whichhas important research significance for AUV to operate long time tasks.In this paper, the following sections are studied：Firstly, system model of various parts of the navigation system are built, includingdetection model of mechanical scanning imaging sonar; the measurement model of linearfeature, Doppler and compass; the model of static characteristics. Finally, the AUV motionmodel is proposed, and the relation between the coordinate is described.Secondly, through the analysis of the data format of mechanical scanning imaging sonar,completing the pretreatment of the data. An improved hough transform algorithm isintroduced for line feature extraction based on the traditional hough transform. An experimentusing the sea trial data is taken out to test the improved algorithm. The result shows that theimproved algorithm can effectively extract the features of the structyred port.Thirdly, the EKF-based AUV navigation method is studied. The basic principle of deadreckoning and its problems are given. After analysising the principle of EKF, the EKF-basedalgerithm is designed and the steps of it ars described.Finally, the proposed algorithm is validated using sea trail data. The result of EKF-basedalgorithm is compared with dead reckoning, and the error curves are analyzed. Experimentalresult shows that the EKF-based navigation algorithm can obtain higher positioning accuracy,which can meet the requirements of long time operation.