Study On Robot Teaching-playback Method Based On Computer Vision

Nowadays,industrial robots should adapt quickly to the changes in tasks with the higher flexibility of the production line.This makes the robot teaching complex and cumbersome.So the traditional method can't fulfill such requirements.The position and posture of index finger fingertips were estimated by computer vision technology.And the teaching of the robot can be achieved with coordinate transformation.This method is simple,easy and general.This improving the usability of the robot.The main contents of this paper are as follows.Firstly,the histogram of the image in HSV color space is used as a description of the target model and the candidate model.The similarity of the two models is calculate by BhatTacharyya distance.The real-time tracking of the human hand is achieved based on particle filtering algorithm and MeanShift.The accuracy of the former is high but the real-time performance is poor in motion tracking.The latter has good real-time performance but poor accuracy.Moreover it doesn't work when the target fast-moving and the direction changes rapidly.In this paper,the internal parameters of Kinect depth and color cameras were calibrated.And the depth information is used to calculate the position of the palm in the camera coordinate system.A color Mark point is designed to solve the correspondence relationship of the target feature point between the target and the image coordinate system in target pose estimation.The Dogleg algorithm is used to obtain the position and orientation of palm in the camera coordinate system.In order to establish the position relationship between the index finger and the palmar,this paper establishes the kinematic model of the finger,and uses the model-based PSO optimization method to estimate the joint angle of the index finger.The contour points of fingers are extracted and segmented.A method for computing the similarity of finger joint angles using curvature integral is proposed as the objective function of PSO algorithm.Aim to prevent the algorithm from falling into the local optimal solution,the particles are mutated in the update phase.The position and orientation of the index finger are measured and transformed into the task path with Coordinate transformation.The median filter has been used to make the task path smoother.A vision-based robot teaching system software is designed,and its effectiveness is verified by experiments.