자율 수중 로봇을 위한 크기 변화에 강인한 영상 기반 물체 인식

Abstract

This paper introduces vision-based object detection techniques for autonomous underwater robots. The accurate localization is very essential for the complete accomplishment of underwater task. But the sensors or environments for the accurate localization are hardly provided. So, in the previous work, we proposed the weighted template matching techniques for vision-based object detection method using a camera in the structured environment. However, the performance was not satisfactory in case the size of object is varying. In this paper, the object detection method has been enhanced to be robust in scale change by employing color-based segmentation. For the segmentation techniques, we assumed as the deep-sea environment that the light source installed on the AUV is dominant and the hue and saturation of the artificial landmark is higher than those of environment. Using the assumptions, the regions of high hue and saturation have been extracted, and some candidate regions have been selected depending on the size and aspect ratio. The selected candidate regions have been scaled as template size and classified relying on the correlation coefficients calculated by weighted template matching technique. The performance of the proposed algorithm has been demonstrated through the water basin experiments using an underwater robot platform yShark made by KIOST.urate localization are hardly provided. So, in the previous work, we proposed the weighted template matching techniques for vision-based object detection method using a camera in the structured environment. However, the performance was not satisfactory in case the size of object is varying. In this paper, the object detection method has been enhanced to be robust in scale change by employing color-based segmentation. For the segmentation techniques, we assumed as the deep-sea environment that the light source installed on the AUV is dominant and the hue and saturation of the artifi