The Research Of Soccer Robot Vision System

Soccer robot is a variety of disciplines and technologies in an integrated product, whichitself is a typical nonlinear control object provides a very good experimental platform for avariety of control theory.FIRA series of soccer robot is composed of four parts of the visual subsystem,decision-making subsystem, communication systems, robot car system. This paperdescribes a series of soccer of FIRA Robot Vision subsystem design and research.The first detailed description of the basic Technical principles of visual computing theory andwell-known the Marr visual computing theory. Robot vision system, image input, imagepreprocessing, the initial human-computer exchange, image analysis, the result output of thisfive-step process followed by the description.In the process of image preprocessing, the BP neural network-based image compressionalgorithm to reduce the image extent of contamination in the environment and to reduceredundant information in the image. Also proposed image correction algorithm based on BPnetwork, the algorithm correction due to the lens to produce nonlinear distortion obvious than theeffect of geometric distortion of the image correction algorithm.Image analysis step, gives mc color image partition method of sample Threshold value,which is a feature clustering method. Through analyzing the samples in the color space, buildsthe object sample’s superset in color space, describes the result of the multi-object clustering withthe eigenfunction in color space, and finally, stores the result of the eigenfunction into the objectcharacteristic repository. In the process of the real-time image partition, using the eigenfunctionto realize the partition of image coordinate set is very convenient and easy.Finally, design a new color scheme and identification algorithm to improve the recognitionspeed. Color-coded design has always been a core technology in the visual system, the color isgood or bad, the pros and cons of the color search and algorithm performance level has asignificant impact on the visual system performance and efficiency. Traditional two color schemehas shortcomings, or is easy to environmental impact or by its own noise interference, making iteasy to identify error, or the algorithm is very complex, take longer to run, cause the system toreal-time low and the overall performance reduced. The new color collection of the advantagesof two traditional programs, namely, direct access to the center and to reduce the interferencecaused by recognition errors due to the environment. The experimental test platform to test newcolor scheme greatly reduces the occurrence of the phenomenon of adhesion, and effectivelyreduce the error recognition rate, while the algorithm time complexity is smaller. The accuracy ofaccess to information captured images of poor quality and a greater impact, the differencebetween the rate of this new color scheme algorithm of the camera is very demanding, but on thewhole is also desirable.