Research On Machine Vision Based Precise Assembly For Miniature Parts

In manufacturing of precise miniature products and devices, the size of parts to be assembled is generally from sub millimeter to tens of millimeters, and high accuracy are require. At present, assembly is generally carried out manually under the microscope, which results in lower yield percentage and poor uniformity of product, for the workers, they have to be trained for better skills. With more requirements on quality stability and yield of the product, manual assembly is difficult to satisfy these requirements. Machine vision-based assembly system shows the advantage in improving precision, guaranteeing the stable quality in production and increasing production efficiency. In the vision-based assembly system, the major task of machine vision is to implement identification and location of the target micro-objects; it should be able to obtain the distance, shape, position and orientation of the micro-parts under certain requirement of precision. Oriented to automatization and batch production of precise and complex miniature parts, for improving quality and uniformity of product, an assembly system based on machine vision for miniature parts has been developed. The assembly system solved the typical problems in the assembly process, including the great difference in size, shape and surface feature among parts; tolerance required for both position and orientation etc.The precise assembly system consists of four functional subsystems, including computer control subsystem, vision subsystem, robot manipulator subsystem and rotary worktable subsystem. Vision subsystem is designed and complete system control software is developed. Vision subsystem includes three DFO motion platform, industrial digital camera and telecentric lens. Firstly, vision subsystem acquire image of parts and then preprocess the image, including smoothing filtering, image enhancement, binary segmentation, mathematical morphology and image mosaic. Secondly, system recognize the target parts by using the methods of edge detection, contour extraction and template matching and obtain the position and orientation of the parts by using the methods of minimum bounding rectangle, centroid and principal axis. Finally, robot manipulator is under control with the motion information through transformation of special coordinates and then accomplishes the precise assembly task. Experiment shows that the assembly system improves uniformity of products and production efficiency, lower skill requirement and work intensity of the worker as well as achieves the accuracy requirements.