Research On Batch Core Appearance Defect Detection Algorithms Based On Ten Million Pixels

The magnetic core is an important part of the chip inductor.In order to soldering the enameled wire ends,the silver plating layer is needed on both ends of the core.However,there may be deviation or damage to the magnetic core during the spraying process,which will cause adverse effects on the quality of the chip inductor.In this paper,tens of pixels of the working core are important parts of the chip inductor.In order to soldering the enameled wire ends,the silver plating layer is needed on both ends of the magnetic core.However,there may be deviation or damage to the magnetic core during the spraying process,which will cause adverse effects on the quality of the chip inductor.In this paper,thousands of pixel industrial cameras are used to acquire thousands of core images in the whole fixture,and a magnetic core defect detection algorithm is designed to detect defects in magnetic cores.The flow chart of magnetic core defect detection algorithm is given.A single core image is extracted from the thousands of core images in the fixture by row and column segmentation algorithm,and the tilt correction of magnetic core image is realized by using the image rotation algorithm.A magnetic core flaw detection algorithm based on neural network classifier is designed,which completes the magnetic core missing corner with high detection probability and low false detection probability.Detection of trap.By extracting the edge,black edge,dirty and fuzzy feature vectors of the magnetic core image pixels,the detection algorithm of edge defect,black edge defect,dirty defect and fuzzy defect is designed to achieve the classification of magnetic core and non good quality products.Through the test of magnetic core image samples,the algorithm has good detection probability(94.87%)and low false detection probability(1.65%),which achieves the design requirements of the system.