A Practical ROI and Object Detection Method for Vision Robot

Abstract

For visual robots, accurate recognition of objects and extraction of useful features are the basis for the next action. In order to exclude the influence of non-critical regions and improve the processing speed, this paper proposes a target detection method based on regions of interest (ROI). Firstly, the preprocessing steps such as filtering, color space conversion, and contour extraction are carried out for the acquired image frames. Then, the contour mask is used to obtain the irregular ROI. After that, color, shape, area and other information are adopted to identify the target and extract features such as angle and distance. This method achieves excellent performance on monocular robots.