Autonomous target tracking by AUVs using dynamic vision

Proposes a method for underwater target tracking using optical data, which consists of color, texture, shape and dynamic properties of the environment. Here, how the dynamic properties from image sequences can be used for target tracking by autonomous underwater vehicles (AUVs) is studied. Optical flow techniques are used to derive the dynamics of the images. The objects of interest are extracted from the images by using the dynamical properties and their optical features. The consecutive dynamic behavior of objects of interest is estimated based on the current dynamics. Using the predicted dynamics, the region of interest (ROI) in the image can be identified reducing the amount of data to be processed. This increases the speed of processing using the hardware available in small pressure hulls. Image dynamics and feature position information fused with other on-board sensor information, the navigational commands for the AUV are derived. In the paper, the extraction of objects of interest from the images based on dynamic vision is presented and the performance of this method for underwater target tracking is demonstrated by experimental results.