A new image processing method of underwater sonar is proposed. To improve the quality of underwater obstacles' recognition, an adaptive fuzzy clustering algorithm is proposed for underwater sonar images. In order to solve the problems from high dimensional underwater data, the principal component analysis (PCA) is introduced for lower space data. The results of experiments demonstrate the performance and validate the effectiveness.
Underwater vehicles are now playing major roles underwater missions. They can be applied to perform complex underwater tasks, such as the offshore, underwater construction work, environmental studies, sea bottom surveys, etc. In order to show underwater environment, necessary information need be collected and sent back to the pilots. Image devices and sensors should be equipped. For example, optical cameras are equipped in several vehicles. However, in practical application, optical devices can hardly provide enough information due to low visibility in the water, especially the turbid water. Acoustic image sonar or acoustic camera is an important device which can provide much feedback sonar information in a long distance, such as ultrasonic pulses including the time of flight and amplitude. It can lead to fast and effective image training in sonar interpretation. These are very necessary for the application of underwater detection, tracking and obstacle avoidance etc (Auran and Silven, 1996, Borenstein and Koren, 1988, Kuc, 1993, Kurz, 2004, Petillot, Ruiz and Lane, 2001, Yu, 2008). In the work time of underwater vehicle, poor quality of sonar image data and the boring, long distance easily cause operators to become tired and make mistakes. In the real applications, obstacles are usually considered as big problems for underwater vehicle. In order to avoid the manual operation involved in dangerous tasks and to provide the pilots with taking decisions about manipulation and navigation task, autonomous or automatic systems are taken into account.