Machine Vision Based Precise Measurement And Assembly For Miniature Components

The demand for highly-integrated micro electromechanical systems is increasing dramatically, but their integration process still depends on the experience of professional workers, which makes the assembly stability and accuracy not highly met and the process of micro products industrialization slow down. Due to the advantage of machine vision technology, the computer vision based assembly system for miniature components is widely put into industrial manufacturing use. As a result, the assembly stability and accuracy are greatly improved while the working intensity is effectively controlled. So computer vision based assembly technology is attracting more and more researchers'attention and becoming very important technology innovation aspect of industry modernization.The key issue for machine vision based assembly technology lies in the achieved recognition and position accuracy of machine vision system and the consistency of whole working platform. The parts to be assembled always have the problem of poor surface condition, wide range of dimensional span and varied manufacturing shapes. They can not only influence the parts edge detection accuracy but also raise high demand on the generality of assembly system. In addition, the position uncertainty caused by movement of relevant functional subsystem is also a very essential problem that needs to be further discussed during system construction.In order to solve the micro assembly difficulties mentioned above, a double camera auto-assembly system for multiple parts is constructed and put into use. Double camera is applied to meet the different demand of detection vision field. Multi-functional manipulating mechanism, which includes both mechanical gripper and vacuum pad, is designed to ensure the flexible clamping of different fragile assembly parts. The real-time monitoring strategy is carried out, which uses the feedback to control system's next motion.After proper illumination and image preprocessing is implemented, the advanced image process is used to extract the target contour. It includes canny detection algorithm, global points scanning method and dynamic centroid position scanning method. Afterwards different kinds of algorithms are carried out to fulfill image matching and positioning, which is consisted of image mosaic positioning method, Minimum bounding rectangle algorithm positioning method, least square fitting method for line and circle and randon angle detection method. Frame transformation, the parameters of which are obtained through camera calibration and system calibration experiments, is applied to calculate the displacement for aligning and assembling task later. Besides, an indirect vision measuring strategy is developed to fulfill the assembly of certain parts which can not acquire real-time image detection and position.The experiment results demonstrate that high assembly accuracy is achieved while assembly efficiency and stability is improved. Consequently, this promises a much higher yielding rate and economic profits.