Tracking of objects in the images received from sector scan sonar is a very important requirement for Autonomous Underwater Vehicles (AUVs) in order to avoid the obstacles that may come in its path during its missions. This involves proper segmentation where the image is segmented into objects, shadows and sea bottom reverberation regions, followed by extraction of the edges of the objects. This then leads to identification of the objects of interest. The objects detected in the images received from the sonar are then tracked and their trajectories are determined. These are then correlated with the speed and direction of the AUV. Subsequently, by using the triangulation method, the collision course on which the collision is expected to occur is calculated. Thereafter, by giving the suitable command to the AUV, the collision is avoided. In this paper, segmentation, extraction of objects, tracking based on the centroids of the objects, along with the calculation of collision course, have been presented. The calculation of the trajectory has been done through the implementation of the Kalman filter, which has been undertaken in MATLAB.

The images in sector scan sonar are produced by a sensor array which electronically scans a horizontally narrow beam to insonify an arc in a set direction. There are various methods which have been used for tracking objects in the image sequences received from the sonar fitted on Autonomous Underwater Vehicles (AUVs). These methods of tracking the objects are primarily based on the applications and their usage. But, if collision avoidance is the main aim, then the time required for calculation becomes an important criteria as the action has to be taken before the collision occurs. Towards achieving this, the following steps are required to be undertaken.

Kalman filter is one of the well-known and often used tools for stochastic estimation of variables of interest from noisy sensor measurements. Kalman filter is simply an optimal recursive data preprocessing algorithm and is a set of mathematical equations that provides an efficient computational (recursive) means to estimate the state of a process in such a way that it minimizes the mean of the squared error (optimal). It applies to stationary as well as nonstationary environments. The solution is recursive, which means that in each updated estimate, the state is computed from the previous estimate and the new input data, such that only the previous estimate requires storage. The filter is very powerful in several aspects: it supports estimations of past, present, and even future states; and it can do so even when the precise nature of the modeled system is unknown.

Telecommunications Journal, Image Processing, Kalman Filtering, Centroid-Based Tracking, Autonomous Underwater Vehicles, AUVs, Tracking Algorithm, Rreverberation Regions, Ccorrector Equations, sstochastic Estimation, Pprocess Model, Measurement Model, Squared Error, Predictor Equations, Predictor-Corrector Algorithm.