Real-Time Reliability Evaluation of Optical and Acoustic Images for Feature-Based Localization of AUV

We propose a method of evaluating reliability of optical image and acoustic image for localization of AUV in real time. Some studies about SLAM using vision or an imaging sonar have been suggested, but both sensors have their own advantageous cases. Using proposed method, AUV switches the utilizing sensor in appropriate case in real time, and always navigates with the most useful sensor. We defined each reliability value of optical and acoustic images using the number of features. We acquired image pairs in seawater and water tank with AUV Cyclops. Using linear discriminant analysis, the parameters to compare their reliability values were found. We compared the results depending on the experiment conditions, and then checked the processing time.