Research On The Core Algorithm Of PLCC-type Component Based On Visual Inspection

PLCC(Leaded Chip Carrier Plastic) is a common type of package in the surface mount process. In the process of its production and placement, the visual inspection is gradually becoming the main strategy of the detection because of the good real-time, high precision and fast speed. Core algorithm research on visual inspection of PLCC components not only can be used for the production quality detection of PLCC components and the component vision detection in high precision SMT machine, also can make algorithm reserves for slender rectangular pin components detection. In this paper, the core algorithm is divided into three parts: preprocessing, parameter acquisition, pose and defect detection. Selecting algorithm and designing flow to make the test results meet the requirements of the detection system, and the specific research contents are as follows:Firstly, the preprocessing stage is divided into two processes of image binarization and contour tracing extraction. According to the characteristics of the image gray histogram, four binary segmentation algorithm are chosed to segmente the component pins, and the improved algorithm of each algorithm are studied, according to the comparison of segmentation results and real-time, the paper determines the one-dimensional fast Otsu segmentation as the pretreatment of the binarization algorithm. Two kinds of contour extraction algorithm based on topology theory are carried on the test and analysis, and the extracted algorithm of the outermost contour only is selected as the pin contour tracking algorithm.Secondly, the parameters information is divided into component parameters, pin parameters, and pin group parameters, and a kind of parameter acquisition process without depending on the prior data is determined. Aiming at the realization of the key algorithm of the minimum area bounding rectangle extraction and pin clustering are analyzed and discussed, using the improved rotation method to abtain the minimum area bounding rectangle of the convex hull of the pin contour and the k-means clustering algorithm with initial center point improving to cluster the pin. In addition, a kind of pin classification algorithm process is designed. The accuracy of the acquired pin parameters by the method this paper proposed is statistically analyzed, and tests show that the parameter acquisition algorithm this paper designed can meet the precision requirements of the system.Thirdly, the component detection is divided into pose detection and defect detection. Combined with the weighted least square algorithm and the fitting rectangle algorithm, four kinds of obtaining pose algorithm by direct calculation are proposed, and the test shows that the accuracy of the fitting rectangle method is the highest. A template matching algorithm with rotation angle template to determine the position is proposed combined with accelerating normalized cross correlation matching algorithm. The component defects is divided into four kinds such as the brightness can’t handle, the component type errors, the component parameter errors, the offset of the component beyond the scope, and gives the detailed introduction. In this paper, the position information obtained by the rectangle fitting method and template matching method is tested and analysied, and determines the rectangle fitting method as the pose detection algorithm, and gives the overall algorithm process containing defects detection to inspect PLCC component.