In order to improve the automation level and production efficiency of auto-assembly line, an auto-grasping and assembly system is proposed based on machine vision to replace manual sorting and assembly. Camera calibration was carried out by OpenCV for parameters inside and out. Based on One Dimensional Hough Probability Transform （1DPHT） algorithm, the pose of target was solved. And the Kalman Filter (KF) algorithm was used for tracking and location forecast for moving target. The visual tracing and grasping platform based on Selective Compliance Assembly Robot Arm（SCARA）and auto-assembly platform based on 6R robot was built for experiment. It shows that the system proposed is valid in auto-grasping and assembly in two dimensional (2D) space with preferable efficiency and accuracy.