Study On Identification And Tracing Red Ball And Self-localization About NAO In RoboCup

Robot technology embodies a national high technology level and industrial modernization degree, is a comprehensive application technology. Soccer robot highly integrated automation, machinery, sensor technology, computer, artificial intelligence and other multi-disciplinary knowledge and research, is one of hot research direction of today’s science, which represents the forefront of cutting-edge technology development around the world. RoboCup robot soccer tournament is one of the high-tech widespread international confrontation activities, NAO fully autonomous soccer robot’s sensors, optical-system and controller etc, all kinds of component integration, is the most widely used worldwide in the academic field of humanoid robots.According to the rules of the NAO robot soccer match in the RoboCup, NAO robot is mainly through the camera to get real-time information and by color to identify the pitch of the target object and environmental information. In order to better adapt to the complex real-time environment on the pitch, requires that the robot’s image processing algorithm has high real-time performance and accuracy. During the race, NAO robot through visual system finding the target red ball, and quickly trace the red ball is a foundation. After find the red ball robot through find goal, adjust the position, select the ball or shooting, finally finished the race.NAO robots only know exactly what their position, and how to reach destinations from one location to another, can effectively perform specific tasks. So the robot self-localization during the game is a very important task, also a very crucial technology. Monte Carlo particle filter positioning algorithm has been widely used in the positioning of the robot, while many of the improved algorithms can improve the efficiency of the Monte Carlo algorithm. Based on NAO robot as the research platform, this paper mainly studies the standard platform league (SPL) game red ball target tracking and identification and robot self-localization of Monte Carlo particle filter algorithm. The specific work this paper studies as follows:1. This paper analyzes the advantages and disadvantages of RGB color space, decided to adopt HSV space as the color of the visual system, it can weaken the scene illumination change on the impact of the visual system. After image preprocessing like color segmentation, de-noising, color threshold determination, etc. A simple strategy to identify red ball based on HSV color space is proposed.2. This paper analyzes the monocular distance tracking methods and their advantages and disadvantages, puts forward a method of central vision tracking red ball strategy, namely the NAO robot in the process of tracking target red ball always makes the target red ball in the center of their field of vision, which is more suitable in actual scene application compared with the range distance finding strategy.3. Monte Carlo particle filter algorithm is an iterative algorithm based on probability calculation, can easily solve the non-gaussian problem caused by robot "kidnapping", can achieve global self-localization work effectively, and many improved Monte Carlo particle filter algorithm to reduce the computational cost of filtering algorithm, and improves its stability. Compared with Extended Kalman filter algorithm of the overall performance, get the Monte Carlo particle filter algorithm is compared with other filtering algorithm is more suitable for solving the problem of nonlinear and non-gaussian system, namely is suitable for solving the robot self-localization problem.