Image Segmentation And Target Detection In Robot Vision System

Vision system is an important part in a robot system. Functionally, a robot vision system contains the following processes: the obtaining of digital image, the image pre-processing, character extraction and image identification. To obtain digital image, the environment around robot is captured by optical lens and then digitized into a digital image; then the digital image will be adjusted by using geometrical transform and image transform, which will be easier to be identified and processed, and some important characters will be abstracted; finally the vision system find out the target based on the processed image.Based on existing technology in image segmentation and identification, this thesis compared some common used algorithms, and proposed some improved algorithms. The main contributions include:1. Review the development of image segmentation and character extraction in image process and particularly in robot soccer system.2. For the application environment of robot soccer system, and to meet the need of the robot soccer system, this thesis analyzed some different color spaces, and selected YUV color space as the work space of the vision system, in which the disturb of the change of illumination is minimized.3. To meet the real-time demand in the robot soccer competition, this thesis proposed a color image segmentation method using improved threshold vector algorithm in YUV color space. The new algorithm can classify 8 colors in one time, so the speed of the algorithm is improved a lot.4. Most existing identification algorithms used in robot soccer vision system need to scan the image another time after all the pixels' color have been classified, which is time consuming. In order to save time, a seed-fill algorithm based on grid search is introduced in this paper. The new algorithm is much faster.5. After compared some different types of color label in robot soccer system, which have different effects on vision identification. This thesis designed a new color label. In the new color label, the angle identification only depends on TeamColor, and the new color label is easier to be identified.6. Applied the improved color image segmentation method to human face detection in color image.