AN ADAPTIVE VISION-BASED SENSOR FOR UNDERWATER LINE DETECTION EMPLOYING SHAPE AND COLOR IMAGE SEGMENTATION

JORDAN, MARIO; TRABES, EMANUEL

Hong Komg

Congreso; 3rd International Conference on Control, Robotics and Informatics (ICCRI 2014); 2014

Science and Engineering Institute (SCIEI) Hong Kong

This work aims the design and implementation of an adaptive visionbasedsensor for detecting a pipe on underwater scenes in real time. The motivationis focused to future applications of vision servo control in underwater vehicles.The approach employs color and shape image segmentation together with anadjust mechanism that aims continuously in time to reach the best setup of aparameter set of the color image segmentation. The sensor performs very welleven in the case of large and rapid changes in the scene illumination. On the basisof many experiments carried out in real scenes and the comparison with similaralgorithms in the state-of-the-art field on the same application, the approach gets abetter positioning with respect to related results above all in the case of extremelychanging and poor luminance conditions. As drawback, the required computationtime to achieve optimal values for the first time (auto-tuning phase) may be large;contrary to the adaptive ongoing process, in where the optimization is much moreagile.