An Improved Otsu Threshold Segmentation Method for Underwater Simultaneous Localization and Mapping-Based Navigation

Abstract

The main focus of this paper is on extracting features with SOund Navigation And Ranging
(SONAR) sensing for further underwater landmark-based Simultaneous Localization and Mapping
(SLAM). According to the characteristics of sonar images, in this paper, an improved Otsu threshold
segmentation method (TSM) has been developed for feature detection. In combination with a
contour detection algorithm, the foreground objects, although presenting different feature shapes,
are separated much faster and more precisely than by other segmentation methods. Tests have
been made with side-scan sonar (SSS) and forward-looking sonar (FLS) images in comparison
with other four TSMs, namely the traditional Otsu method, the local TSM, the iterative TSM and
the maximum entropy TSM. For all the sonar images presented in this work, the computational
time of the improved Otsu TSM is much lower than that of the maximum entropy TSM, which
achieves the highest segmentation precision among the four above mentioned TSMs. As a result
of the segmentations, the centroids of the main extracted regions have been computed to represent
point landmarks which can be used for navigation, e.g., with the help of an Augmented Extended
Kalman Filter (AEKF)-based SLAM algorithm. The AEKF-SLAM approach is a recursive and iterative
estimation-update process, which besides a prediction and an update stage (as in classical Extended
Kalman Filter (EKF)), includes an augmentation stage. During navigation, the robot localizes the
centroids of different segments of features in sonar images, which are detected by our improved Otsu
TSM, as point landmarks. Using them with the AEKF achieves more accurate and robust estimations
of the robot pose and the landmark positions, than with those detected by the maximum entropy TSM.
Together with the landmarks identified by the proposed segmentation algorithm, the AEKF-SLAM
has achieved reliable detection of cycles in the map and consistent map update on loop closure, which
is shown in simulated experiments.