Research On Marked Line And Ball Recognition Applied To Game System Of Middle Size League Soccer Robot

Robo Cup, as one of the most influential in the world of two robot soccer worldwide, mainly involving cutting-edge research and technology integration of artificial intelligence, robotics, vision sensors, communications, precision machinery and other fields, is a high-level, high-tech game. With the continuous development of science and technology level, many country has paid more and more attention on robot reseearch. For medium-sized group of soccer robots, the research includes many aspects, such as hardware design, visual design, motion control, path planning, with offensive and defensive strategies and so on. Among them, the visual system as a robot’s eyes, is the basis of the robot can successfully complete the task. The study of visual recognition of soccer robot can expand the application fields of machine vision, enrich and develop the theoretical knowledge of machine vision and image processing.Medium-sized group of soccer robot vision systems, mainly through omnidirectional images to complete the collection and identification of targets, based on the collection and analysis of relevant literature, this paper mainly pitch the ball and robust identification marking line for doing research. the main research are as follows:(1) recognition of marked lineImage preprocessing part, the common method for image smoothing median filter, mean filter and Gaussian filter to analyze, select, and then the image edge detection, select the Gaussian filter with the Laplacian operator to use, Then the traditional Hough transform to improve, the first image in the neighboring pixels to detect the clustering of similar pixels to form a collection of connected pixels, and then use the random Hough transform detection to accurately identify the corresponding image segment.(2) recognition of ballSince the height of the color coding of the game environment, this paper from the color of the ball to start the identification. First, by comparing the color model used to select the HSV color model is more in line with the human visual characteristics, and then on the shortcomings of traditional BP algorithm for analysis, an improved BP neural network algorithm, using its good classification results to adapt to the environment changes in illumination, to make an accurate identification of the target.At the end of this paper, it summarizes the main research of this task, also points out the shortcomings and the problem which needs further study in the future.