A Robust Approach of Sonar Image Feature Detection and Matching
Shoudong Shi, Demin Xu · 2009
This paper is concerned with Modulus Maximum of Wavelet Transform (MMWT) and a graph theoretic method. The methods are applicable to extracting features of seafloor sonar image and data association problems. We will first get imagepsilas modulus and moduluspsila direction matrix by MMWT method. And according to calculating moduluspsila threshold, obtain the geometric features of the image or the point features. Then calculate geometric centrobaric coordinate of the geometric features as matching point.For point feature, featurepsilas Vector will be created by the combination of region direction of modulus. For geometric feature, featurepsilas Vector is its perimeter and area information. At last, the key points between images will be associated by Maximum Common Subgraph method and validated by the feature vectors. The experimental results show that the methods are reliable and robust in continuous sonar image of seafloor.