Font Size: a A A

Research And Design Of Humanoid Soccer Robot Vision And Decision-making System

Posted on:2013-06-27Degree:MasterType:Thesis
Country:ChinaCandidate:Y LiuFull Text:PDF
GTID:2268330425491870Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
This research on soccer robot has obtained rapid development in recent years, which makes this muti-disciplinary overlapping topic recevice more and more attention. The object of this thesis is the RoboCup standard platform robot which contains vision system, deci-sion-making system, communication system and so on. It emphatically studies and designs the vision and decision-making system of the robot.The vision system must have real-time, accuracy and robustness.In this thesis; the vision subsystem uses a COMS camera as the image acquisition device. In vision system, the image horizontal is calculated, which can reduce the number of pixels to be processed. With the re-quirements of the tasks of the visual system, a colortable is established based on kd-tree. And a method by using a grid with variable step is used to split the image to establish the segmen-tation line. Depending on the relationship among the segmentations, the target region can be rebuided. In the end, according to the different requirements, the targets can be classified. In addition, the identification of the football game scene for the different light conditions, this thesis proposes a method based on a threshold of two-dimensional space in YUV colorspace, and the method has better adaptability to different light conditions.Decision-making system is a key step in the football match, it must meet the effective-ness and real-time requirements. Therefore, the models and algorithms of the decision-making system must satisfy the high demands. With a lot of decision-making tasks, a automatic state machine is used to manage the behaviors of the robots. Using the information provided by vision system, a robot soccer shooting algorithm is discussed in detail, and a fast shooting al-gorithm is presented. Combined with the shooting algorithm, an optimal motion path has been given. Finally, the dynamic role assignment strategy has been discussed.In the end, this thesis discusses system’s defects and deficiencies. In the same time, the future development foreground of the RoboCup standard platform has also been prospected.
Keywords/Search Tags:Nao, Motion imgae, colorTable, region segmentation, region construction andrecognition, shooting algorithm, path planning
PDF Full Text Request
Related items