Research On Image Processing And Recognition Techonology For Chip Mounter On-the-fly System

Surface mount technology (SMT) is one of the most advanced electronic assembly technology whose occurrence makes a revolutionary change in the electronic assembly industry. With the rapid development of microelectronic industry, the demand of chip mounter, which is the most complicated core equipment in the SMT production line, is growing faster and faster. Having the advantages of high efficiency, non-contact, high reliability and low cost, imaging processing and recognition technology (IPRT) has been widely used in the aspects of SMD components detection, nozzle detection and PCB mark detection in modern chip mounter systems. With improvement of chip mounter’s performance, IPRT is playing an increasingly important role in the area of SMT.Due to the ambient interference when obtaining the images, which are unavoidable corrupted by noises, it is important to perform an image preprocess procedure in order to suppress the effect of noises to the following recognition algorithms. Such preprocessing procedure consists of three major steps. Taking into account the computational load, the first step uses a Gaussian filter to remove the gray image noise. An iteration based binarization operation is then employed in the second step according to empirical evidences.The output of the binarization process is then fed into the third step, which purports to repair the binarized image using a denoising technique based on mathematical morphology.To accurately recognize the SMD components, we employ a minimum enclosing rectangle algorithm to identify the center position of the SMD components whose rectangular geometry characteristics are exploited. Rotation angles of the SMD components are also determined with a line fitting algorithm based on the subpixel technology.As for the identification of the nozzle’s center, we adopt a least square circle fitting method that makes use of the circular geometry features of the nozzle. Due to errors of manufacture and assembly, the nozzle is not exactly perpendicular to the X-Y mount plane. This will cause the position change of the centers of SMD components when the nozzle rotates to correct the angle of SMD components, and thus affect severely the final mount accuracy. To tackle this problem, we construct a correction model based on the concepts of "revolution" and "rotation". Based on this model, we further propose a nozzle correction algorithm based on image recognition and circle fitting. By tracking the center position of the nozzle when it is rotating, this algorithm can achieve the purpose of correction satisfactorily.Besides the above-mentioned work, we also investigate one of the main tasks of the placement machine vision system-the recognition and positioning of PCB mark, which connects the coordinate system between that of SMT system and PCB board. Simulation experiments are conducted to verify the effectiveness of template matching method and circle fitting method based on the circular geometry of the PCB mark.