On-the-fly Vision Based Detection Algorithm And System Development For Multi-nozzle Chip Mounter

Machine vision system is one of the key technologies of high-performance chip mounter. It plays an important role in increasing speed and accuracy. Aiming at fast acquisition and detection of the components and marks for the multi-nozzle chip mounters, we proposed a multi-component simultaneous acquisition and detection method and a fast, accurate mark location algorithm in this paper. The methods are implemented and used to increasing the visual detection efficiency. The main research contents are described as follows:1. The domestic and overseas research status and development tendency of chip mounter and vision system is investigated in depth. The gap between domestic and overseas chip mounter is analyzed and the research challenges and significance of high speed multi-nozzle chip mounter are necessary.2. Traditional chip mounter on-the-fly vision system can only simultaneously process a few components, in order to increase efficiency, a method of multi-component image synchronous acquisition is proposed. The corresponding on-the-fly vision system is defined based on the selection of required industrial cameras, lens, illuminant and so on. Thus, the multi-component partial scan image can be acquired clearly. In addition, the planar calibration boards and calibration method are designed for the vision system to make a fast and accurate camera calibration.3. In order to accelerate the speed of image processing, an automatic extraction method for edge region of component and PCB mark is proposed through the combination of binary morphology and image reductiond. Due to the shortcoming of edge extraction algorithm based on two-dimensional image, a fast sub-pixel edge extraction method is proposed though one-dimensional curve fitting based on the pixel edge acquired.4. Considering of various componenrts of surface mount, we divided these components into three categories:leadless components, ball lead components and extended lead components, to improve the universality and flexibility of the detection algorithm proposed. For the leadless components, a weighted least square rectangle fitting algorithm is proposed based on Hessian normal form of line and Tukey weight function to fit the multi-component sub-pixel edges. This algorithm can reduce the impact of outliers caused by uneven edges on fitting and detect the center and deflection angle of component fast and accurately. Considering that the PCB circular mark will become an ellipse shape image, a weighted least square ellipse fitting algorithm is developed based on the algebraic distance and Tukey weight function to fit the mark sub-pixel edge. This algorithm can reduce the impact of outliers caused by defiled edge on fitting and improve fitting accuracy, reduce fitting time, locate the mark center fast and accurately.5. Based on the gantry precision motion stage, the on-the-fly vision system for multi-nozzle chip mounter is developed though the hybrid programming of HALCON and Visual C++. Furthermore, the system is implemented and verified though the experiment tests on various muliti-componrnts and marks.