Research On Localization And Football Detection Of RoboCup Middle-Size Robot

With the rapid development of computer vision technology and the enlargement of its application field, the technology of robot localization and detection technology for football are becoming the research hotspot in the field of medium-sized group of robots of RoboCup.Target detection, localization and trajectory prediction based on the full use of feature information of image is regarded as not only a important theory meaning and practice value for the follow-up functional development of the robot. Therefore, this paper further analysis and research the hot issues under the existing experimental conditions. And provide a method of detection and localization with high accuracy and good real-time for medium-sized group of robots of RoboCup. These research results of this paper are summarized as follows:(1) In this paper, a localization method of geometrical map matching through the white lines of robot football competition venues. First, through the calibration of all-directional visual distance to overcome the error of positioning accuracy because of manufacture errors and installation errors. Then use the Hough transform to detect the information of white line of venues, and determine the properties of the white line through the method of geometric map matching in order to achieve the robot self-localization. Finally, through the coordinate transformation to determine the posture of football.(2) A correction algorithm base on the analysis of the reasons for radial distortion panoramic view of the ball was proposed in this paper, and noise processing for the image has been done in order to reduce unnecessary interference of edge detection. Then through compared and analyzed several existing operators on the edge of the image processing results of edge detection,the author find out the most suitable algorithm of the edge detection operator is Sobel operator. Finally, mathematical morphology processing was done for the edge detected in order to ensure obtain a more reliable and effective edge detection.(3) Base on the analysis of variation rule of each color component values in variable illumination conditions, a method that through the overall image brightness adjusting the threshold of the color component range was build up. So, the stability of color feature extraction in an environment with uneven changes of illumination in playing field was ensured.(4) As the football detection based on a single color feature is vulnerable to interference by the external environment.Therefore, a combination of detection of football including color detection and edge detection was established. When the color test fails, we can identify the football with the information of football edge feature. Consequently, the reliability of detection of football becoming more reliable. Finally, we can predict the trajectory of football and improve the real-time of football detection through geometric reasoning and mathematical calculations to the moving information of football.