Research And Implementation Of Humanoid Soccer Robot’s Vision System

Robot soccer provides a standard platform for researchers to study the related technology, which is a landmark and challenging academic subject, together with interest and participation, therefor obtains the global scholars and robot lovers’ attention. So robot soccer is the study hotspot in recent years, which promotes the robotics, machine vision, image processing and pattern recognition, intelligent systems and other related areas to develop. For humanoid robots, it is very important for the development of intelligent robots to have a similar visual system to human’s.Especially to humanoid robot soccer in RoboCup, the vision sensor is the only race information sources. Therefore, the humanoid soccer robot’s vision system has important theoretical significance and application value.In this thesis, based on a number of the related literatures consulted and relevant technology, the humanoid soccer robot’s vision system with robustness has been studied and implemented. In order to achieve the requirement of real time and accuracy in the robot soccer game, the problem of illumination effects and similar to object recognition and improvements to of target tracking algorithm’s real time have been studied to meet the requirements of the game. Firstly, robot soccer applications and vision module structures have been finished. Then several key issues were systematically analyzed existing on robot’s vision system. For the effect of illumination changes affecting the recognizing, the traditional color table algorithm has been improved. The adaptive color table segmentation algorithm based on the exponential model has been studied. The cube based on adaptive reference image segmentation algorithm has been used to segment the image. For the target identification problem, the run-length encoding algorithm to the segmented images has been to achieve the region’s searching and merging. Then the genetic algorithm has been used to find the optimal solution. Finally, in order to complete the task of humanoid robot’s tracking, the particle filter based on weighted histogram tracking algorithm has been used. To the problem of poor marketability, combining with integral histogram to achieve particle filter tracking algorithm has good results by carrying out experimental comparison.Experimental results show that the proposed adaptive image segmentation algorithm can effectively achieve the target segmentation and has better adaptability to the illumination change. Using Genetic Algorithms can effectively identify the target object. The experimental results demonstrate the effectiveness of the particle filter algorithm to achieve the target tracking in the field.