The Seabed Image Mosaic System Research Based On Multibeam Side Scan Sonar

With the deepening of ocean exploration, sonar detection technology plays an important role in the ocean detection. As the data of sonar only represent the partial information not the whole topography of the seabed, so it has great limitations for the analysis of the resource distribution in the whole piece of marine and precise positioning with underwater targets. So it is very important to get a complete subsea topography image stitched by multibeam side scan sonar image. The sonar are often subject to noise pollution when reflecting and receiving acoustic signals in the complexed marine environment, and thus affect the sonar images. In addition, when using the sonar to detect the ocean, the tow fish may be effected by the sailed attitude and speed of the ship and the waves, which will lead to the distortion appear of the sonar image, and there exists big difference in information of data between sonar and the real seabed. Sonar images compared with the ordinary optical image less the amount of information, the image grayscale range is narrow, and it is difficult in feature extraction,These problems increase the difficulty of recovery real submarine topographic map. In this paper, the technology of multibeam side scan sonar image stitching will be research.For the sonar data, according to the agreement obtain the sonar image by decoding the data and additional information, and according to the additional information correct the speed and grayscale. Analysis of the causes of noise, study the filtering and decreasing noise algorithm.Research the sonar image registration algorithm in this paper, and for undefonrmed and with small deformed image. An improved algorithm for Harris corner detection is proposed by considering the algorithm structure is easy and there are small amount of calculation in Image registration algorithm of the feature of corner. Describe the feature corner with LBP describing algorithm combining with the theory of circular template. And the centroid is regarded as the main direction of the operator to ensure the directional and scale invariance of matching algorithm. An improved algorithm for SIFT is proposed by considering the high error rate in the feature matching using the corner points as features for large deformation sonar image. Nonlinear filtering principle outstanding image contours, and create a nonlinear filtering pyramid. To ensure that the original SIFT algorithm worked well, and improve the accuracy of registration and the number of feature points of the image.The end of the paper analysis of the common image coordinate transformation model and fusion algorithm, and complete the stitching of the sonar image.