Research on Preprocessing of Color Image for Vision based Mobile Robot Navigation

摘要

There are special requirements on image processing in speed and robustness for vision based mobile robot navigation. But mobile robot captured scene color image?s processing has a high computational complexity and low robustness. To reduce data volume and improve robustness, a novel preprocessing method on color image for mobile robot navigation based on principle of threshold segmentation in HSV color space is proposed. Firstly, color image changed from GRB to HSV. Thinking of H component is instable when saturation level is too low, so the principle for single threshold gray image segmentation be used to choose a threshold T on S and divide S into high saturation and low saturation region, we can use H as segmentation basis in high saturation region and V in low saturation region. Then project H component to high saturation region and V component to low saturation region, H and V is stretched to different scope after projection. So, the S component contains not only color information but also gray information and reduces color image from three-dimentions to two-dimentions. Experimental results show that this method has more significant effects in improving the processing speed and robustness of illumination variation and shadows.