Printed Circuit Board Of The Intelligent Detection System Research

With the miniaturization of electronic components, fine development. Electroniccomponents manufacturing used SMT technology, but use this technologymanufacturing process relatively complex printed circuit boards, solder joints withhidden features. Solder joint inspection becomes a important and difficult. The currenttest method of electrical intelligent detection and optical detection is relatively low, sointelligent and efficient detection system is the development direction of PCBinspection. In order to meet the requirements, this paper based on the theory ofautomatic optical inspection analysis of a printed circuit board of intelligent detectionsystem.At first, this paper use EC1350C type CCD color digital industrial camera inPCB image. And then the projection method is adopted in the horizontal direction andvertical direction respectively traversal and background image to segment foregroundpixels pixel in order to separate the background image and image under test, so as toextract the image under test. Next, we make the image enhancement, smoothing,sharpening to remove noise interference. Specific methods by average filtering andmedian filter to smooth image processing to eliminate foreign interference on the PCB;Use the image gray-scale transform to increase the image contrast enhancement;Using Laplace sharpening method for image sharpening processing in order tohighlight the image detail. Finally using neural network BP algorithm and the LMBPalgorithm of PCB solder joints defects identification respectively. Based on SurfaceEvolver software virtual solder short, tin, tin less, erection, welding, offset six majordefects, and generates the solder joints defects standard samples to train and testneural network, the trained network is used to identify solder joints defects, outputidentification results, and through MATLAB simulation training error curve, twoalgorithms of noise and error curve fitting curve graph.This article focuses on pattern recognition, artificial neural network to solderjoints defects using intelligent identification of classical BP algorithm solder joints defects, due to the BP algorithm is slow convergence speed, convergence speed isrelatively fast improved algorithm namely LMBP algorithm, and through theMATLAB simulation experiment. Results show that the improved algorithm caneffectively improve the accuracy and the misclassification rate of defect detection.