Research On Image Processing And Path Planning For Soccer Robot

Soccer robot as a kind of intelligent robot,combined the competitive sports with artificial intelligence,with comprehensive use of mechanical structure,circuit design,pattern recognition,automatic control,wireless communication,decision making and other disciplines of knowledge,is a good practice platform for studying artificial intelligence.In this thesis taking the small soccer robot as an example,the image segmentation recognition,target trajectory tracking in soccer robot vision,and the path planning of soccer robot have been studied.The main work and its achievements of this thesis are summarized as follows.Firstly,the method for segmenting the objects such as boundary in the color image of soccer robot scene is studied.A color image segmentation method based on PCNN and maximum gray entropy component is proposed.In this method,the color image is converted into HSV color space,the image component with the largest gray level entropy is selected,to be segmented by PCNN according to the maximum Shannon entropy criterion.Simulation experiments verify the effectiveness of the proposed method for color image segmentation and the superiority compared with other methods.The objects in the soccer robot race scene are segmented accurately and completely by this method,which is meaningful for the recognition of obstacles in path planning for soccer robot.Secondly,the tracking method of soccer trajectory is studied.Aiming at the problem of target tracking accuracy,an improved mean shift tracking method is proposed.In this method,the target model and the candidate model are calculated by calculating the LTP texture information on the basis of the wavelet transform to extract the image detail information,and the background weight information is introduced when the target model is calculated.The method can be used to track the trajectory of soccer more accurately,and thus provides the basis for setting the target point in the soccer robot path planning.The simulation experiments show that the improved mean shift tracking method can enhance the stability and accuracy of target tracking dramaticlly.Finally,the path planning method for soccer robot is studied.An environment complexity-oriented method for step length adaptive adjustment based on artificial potential field is presented.In this method,the iteration step size is determined by the inverse relation between the step size and the environmental complexity.By means of the presented method,the dynamic path planning is realized efficiency.The simulation experiments in static and dynamic environment show that the proposed method can reduce the number of iteration for the soccer robot to reach the target point,and therefore provide the foundation for the real-time control of the soccer robot.In this thesis,the simulation experiments are completed by utilizing Matlab software.The simulation experiments also display the feasibility of the proposed methods mentioned above for other-objects of study.