Research On Mosaic Method Of Side Scan Sonar Image

Submarine geomorphology is the key reference for marine detections,side scan sonar is one efficient way to acquire them.Original side scan sonar data can be processed into single stripe image,whose quality can be improved by data acquisition operations and filter processing.For now a single sonar stripe image is not satisfied for the marine research needs,searching for an appropriate method of side scan sonar image mosaic is key appointment.Based on this circumstance this paper conducts following studies:(1)Quality control of side scan sonar image.Through controlling data acquisition operations and filter processing,sonar images quality is improved.Using USM filter to enhance image features then using SURF algorithm to match sonar images,meanwhile choosing mean filter and average filter as comparison.Experiments show that USM filter can improve the accuracy of image registration;(2)Image registration method for side scan sonar image.For solving targets dislocation of geocoding mosaic image,four feature detecting algorithms have been studied.Using Forster algorithm,SUSAN algorithm,Harris algorithm and SURF respectively to extract image features.The results tell that compared with the former three algorithms,the features extracted by SURF algorithm own scale invariance and feature descriptor and stronger robustness;(3)Image mismatching points removing and precise mosaic.After enhancing image features by USM filter,SURF algorithm is used to match sonar images and Random Sample Consensus algorithm is used to remove the false matching point pairs.Finally,images are precisely mosaic by wavelet fusion method.Result shows that SURF algorithm solves targets dislocation problems,image accurate mosaic is realized.