Research On Image Building And Scan Matching Technique Of AUV Forward-looking Sonar

The study of underwater robot technology and level of application is a measure ofa country in the aspect of Marine exploitation and utilization of the important basis ofcomprehensive strength. In the complex underwater environment, the autonomousunderwater vehicle (AUV) to achieve a high level of autonomy, first of all must haveindependent vision system and the self-positioning system. Underwater sonar sensorhas incomparable superiority compared with other devices, so as to becominganimportant part of AUV’s "Sound Visual" system. Since the location problem is refersto the robot on the basis of sense their surroundings to estimate its displacement pathand positioning.Firstly, for the852-type forward-looking imaging sonar used by its workingprinciple and echo information, including specific parameters and data formats. Afterfully understand the information introduced echo from data acquisition, analysis to thevisualization of the process in detail.Secondly, in terms of acoustic image processing to the original, the first analysisof the major problems with the images, which contains more noise and distortion dueto drift. In this paper, taking the characteristics of high noise directly with a singlebeam data integrity methods to maximize the retention of information, and thenenhanced by means of reducing background noise and other image binarization, edgedetection technique combined with the data points that meet the thresholdrequirements for screening, and finally get a clear the exact contours of the image.Then, in order to reduce the image distortion effect on the subsequent Scanmatching, this paper tries to introduce in the process of sonar imaging navigationprediction technology. Using kalman filter update DVL and AHRS devices posetransformation valuation, in order to improve the track is the accuracy of the estimatesfor AUV, and displacement of the robot in the conventional sonar imaging methodinto account, the resulting distortion lower easy to Scan images subsequent registration.Finally, a detailed description of the Scan matching algorithm is the most widelyused ICP principles and characteristics, as used herein, the probability of an iterativeclosest point ICP algorithm is extended in a statistical algorithm, the algorithm isgiven after the process and iterative steps, and through the analysis of simulation datavalidation feasibility of the algorithm.