An image processing approach for determining the relative pose of unmanned underwater vehicles

The use of a light source as a beacon is advantageous for the guidance and control of Unmanned Underwater Vehicles (UUVs). This approach allows a follower UUV to determine its relative pose (position and orientation) using low-cost commercial off the shelf (COTS) hardware (e.g., metal halide light sources). In order to design an effective detector unit for the follower UUV and predict its performance, a simulator program has been developed. The program simulates a light field using hardware and environmental parameters describing the light source, water properties, the detector unit geometry and electronic sensitivity. The simulator allows examination of different 3D detector array shapes of varying sizes (physical dimensions and number of detectors). It is convenient to present simulator output as an image, where each pixel represents the intensity logged by a corresponding detector. These image outputs are evaluated for the development of control algorithms for UUVs. Currently control algorithms assume that the water column is uniform with a background noise of known origin. Considered control algorithms are able to provide guidance based on relative intensity values, where the light field samples on the detector array resembles a Gaussian beam pattern. However, disturbances in the medium (e.g., sediment plume) may cause non-uniform distribution of the scatterers that distort the beam pattern. As a result, the control algorithms could misinterpret the acquired image and direct the follower UUV away from the guiding beam. The probability for such a situation increases with distance as the beam diverges. This paper suggests an alternative approach for the development of UUV control algorithms using calculations of various moments of the image (e.g., local Hessian estimations). This method allows the evaluation of the array performance with different array geometries and a varying number of detector elements.