Research On Target Pose Information Detection For Robot Operation

Intelligent robot as a new industry is changing the world which already occupied homes and workplaces like computer.With the development of intelligent technology and the demand of nuclear accident emergency,robotics tend to replace manpower with completing more complex and dangerous tasks. In order to solve robot's problems of object recognit ion,measurement and independent operation in an unknown environment,article proposes a position information detection method research for robot's operation target.First of all,paper analyzes the topic of subject and made clear that it's object, carrier and application which are pose information detection,robot operation goal and robot fetch.Based on it,the paper can be divided into three parts: robot's working space measurement;job target recognition and segmentation;target pose estimation and relative pose with the end of the robot hand.Secondly, in view of the three parts, the paper has been clear about the research target and research content with three parts. It is:the robot's working space measuring based on KINECT; target identification and segmentat ion based on NCC matching and KNN point clouds clustering; pose estimation and relative position calculating based on PCA and D-H coordinate transformation.Finally,by analyzing the concrete research content,paper derived the innovation about method and technologies conjunction.They are:depth map pretreatment based on monte carlo uncertainty evaluation; the NCC matching optimization combining the illumination correction; KNN point clouds clustering improvement combined with image matching center mark.The experimental results show that:the depth map preprocessing laid the foundation for improving measuring precision of the operation space;HSV lighting component of Gamma correction improves the NCC matching light resistance;center of the target solved the problem of the optimal solution of the K value.All above, the paper use c++ language to realize the algorithms in VS2010 platform combining with Open CV2.4.3 and PCL1.6.0.