An Improved Side Scan Sonar Image Processing Framework for Autonomous Underwater Vehicle Navigation

Abstract

Side-scan sonar (SSS) is placed great expectations to support Autonomous Underwater Vehicle (AUV) in self-localization. Current research focuses on feature-based navigation solutions of which the foundation is data association. However, underwater acoustic image processing technique is still unmatured to produce high-quality SSS image and realize accurate registration. This paper presents an SSS image processing framework implemented in an autonomous workflow to provide real-time navigation information. Canny edge detector-based bottom tracking and improved Retinex-based gray level correction are two key algorithms in it. Canny edge detector takes waterfall image as objective, utilizing the information of not only the single ping but also its surrounding pings, which increases the accuracy and stability of sea bottom line extraction. Improved Retinex with Time Variant Gain (TVG) compensation remedies the shortage of attenuation trend correction. Combined with geometric correction module and registration module, an internally coherent framework is proposed to support AUV navigation.