Automatic Recognition for the Position and Pose of Parts Based on Image Moment Invariants and Neural Network
Shugui Liu · Chuangan jishu xuebao · 2006
The automatic recognition for the position and pose of measured part is a key technology in the inspection system of intelligent coordinate measuring machine.Some virtual images of measured part are formed according to each possible placed mode.And the actual image is captured by the CCD camera.The theory of image moment invariants is applied to image matching.The BP neural network is trained by the moment invariants of virtual images and the corresponding patterns.Using the trained neural network as classifier,the pose of measured part is recognized by the moment invariants of captured image.At the same time,the position and orientation of part figure in the image are calculated by the moment of captured image.Through the coordinate transforming,the position and orientation of measured part in the machine reference frame are gained.The experimental result shows that the method of automatic recognition for the position and pose is intelligent and efficient.