Research On The Algorithm Of TR-Type Chip Detectiong And Location Based On The Machine Vision

The TR-type chip detection and location algorithm is studied deeply in this dissertation. There are many kinds of chips and TR chip is a kind of chips with pins. In order to make a proper mounting movement, the paper gets TR chip detection parameters and makes the defect detection. The paper gets TR chip precise position to calculate the offset angle and central deviation while SMT controls compensation offset. That is to say, the mounting accuracy and speed are affected by chip detection and chip location. Therefore, the study of this paper is important. The main contents of this dissertation are as the follows.A pre preprocess of the camera image is made in this dissertation. The light intensity is judge based on histogram and gray value calculation accurately. A variety of filtering methods are analyzed in this paper. Then an adaptive median filter which can both protect the image edge and remove salt and pepper noise is proposed. For a better image segmentation, this paper proposes a two-dimensional OTSU based on traditional OTSU and a watershed algorithm. Finally, the OTSU algorithm is chosen for a high efficiency.The paper proposes algorithm to access chip information. Analyze the traditional reptile’s principles, for its shortcomings, it results badly. Use eight-neighborhood contour tracking algorithm to extract the contour, taking into account that OTSU algorithm is a global method and can’t meet the local requirements, then filter the contour. The paper proposes a method to achieve chip coarse position based on the minimum bounding rectangle of the convex hull, and uses the affine transformation method to get horizontal chip outline. Use the improved initial centers clustering to segment the upper and the lower portion of the pins. Use vertical projection to mark upper and lower pins and segment the foot and heel of pins based on coordinate. The paper proposes methods to get chip parameters, one based on foot projection, another based on direct calculation, the latter achieves high accuracy.According to the characteristics of TR chips, this paper proposes three methods of positioning chips. The method of image segmentation is based on the foot contour by image segmentation and fit a straight line to achieve position information. The method of edge gray matching thinks edge gray as feature, with a distance transform to express background target distance, and is added a pyramid scaling to improve the efficiency of the algorithm. Edge gradient matching, using SOBEL to extract gradient, thinks edge point gradient as feature to operate. Analyze the offset angle and the pin defects such as missing, the height offset and position offset, then defect detection system is designed for TR-type chip, it is possible to detect the defective chip.Algorithm testing is to test accuracy and stability, the testing system contains algorithm for pin parameters and algorithm for positioning. When the light changes, the pin parameter algorithm’s difference range is within ± 0.1mm, then it basically remains stable. Test three algorithms for positioning, and they can meet the accuracy requirements of ± 0.2 °. Compare time and complexity of the algorithm, and select the method of image segmentation. At the same time the paper tests the stability of the positioning algorithm, it results the algorithm’s stability meets the requirements when the light changes.