Research On Image Processing Techniques In Application Of Soccer Robot

Robot vision system is an important part of robot system. It uses Various techniques of many scientific subjects such as patter recognition, artificial intelligence, computer image processing, etc. to fulfill a function like human eyes to perceive and acquire information from the outside world. As a branch of the machine vision, soccer-robot vision system employing region segmentation by color is crucial to fulfill its real-time, precise and robust performance, it provides the information for the decision-making system. This dissertation is devoted to the research on image processing and related techniques for soccer-robot vision system. It represents special theories and algorithms according to the specialties of color images, proposes a whole solution to achieve the goal of defect detection and segmentation.According to the demands of the vision system, we compare with the popular color space such as RGB, YUV, HSV used in the vision system. Basing the results we adopt two pixel classification methods to label the pixel fast, one is based on hue double thresholds and another is based on the YUV cube. A key to the efficiency of our YUV cube approach is a new method for accomplishing color space thresholding that enables a pixel to be classified into one or more of up to 32 colors using only two logical AND operationsConnected regions are formed by examining the classified samples. This is typically an expensive operation that can severely impact real-time performance. Some common used region merging methods are presented in the dissertation and a method of connected region labeling based on run length encoded image are introduced, it takes only two stages to implement the work. The object information such as position and orientation are extracted from the statistics calculations. Noise will interfere and sometimes object information are lost, we need interframe object tracking and prediction. we discuss it though our system does not currently address this directly. This is an item slated for future inclusion into the vision system.