| With the rapid development of microelectronics industry in recent years, printedcircuit board (PCB) has changes toward miniaturization and high density. Thetraditional depend on manual visual inspection method can not meet the requirementof production quality.In order to ensure printed circuit boards welding qualityevaluation of reliability and stability, reduce artificial assessment of the influence ofstability factor, Many domestic and foreign researchers had researched comparedwidely welding defects in printed circuit boards. Now the urgent need to solve theproblem is that How to use the corresponding image processing recognition methodsand optimization scheme, and detection technology based on visual had been attentionwidely. Detection technology based on visual is mainly realized by the digital imageprocessing and pattern recognition method.In the visual inspection on printed circuit board, IC is an important link onprinted circuit boards detection because of chip pin, spacing of small, weldingaccuracy higher. This paper studies the content is IC chip pin of welding defectdetection on printed circuit boards, including IC chip pin image positioning, chip pinnumber, size and spacing extraction. Detect whether there are bridge even, lack ofwelding between pin according to these parameters.This test is to use CMOS camera artificial collection of pictures, positioning totest chip interest area based on prior knowledge with the picture center positioningmethod. In order to detect image recognition more effectively, the original imagemust be enhanced, the results of experiment is that median filtering has the besttreatment effect on chip pin. In the image segmentation, canny algorithm is the mostideal. In order to detect pin number and width the image must be binary images.Binary image is not very ideal through experiment,so the morphological filter is usedto get clear binary image. Extracting connected domain of the image, judge the pin thenumber and width of the pin. According to these parameters as the prior condition,detect whether happened bridge even, lack of welding between the chip pin throughthe pattern recognition.The paper describes the detection algorithm principle and process detailly, andthe experiment is done stability and repeatability. The experimental results show thatthe algorithms are reliable, satisfy the detection requirements for printed circuit board.The innovation of this paper reflected in the following respects:(1) According to the IC pin welding area location problem, this paper proposes acascade refining method on determining pin welding area. This method firstlydetermine the chip in the general area through the detection of square feature of chip,determining the chip pin welding area based on the distribution features of the chiptube feet quickly and accurately. The results show that the method is effective torealize rapid housekee welding area location. (2) In the pin welding defect recognition, the thinking of sequential similaritydetection and the tree classification design algorithm be used. |
