Study On Machine Vision Defects Detection Method About Injection Molding Products With Polished Surface

Injection molding products are used in consumer electronics products more and morewidely, and whose surface quality is increasingly stringent. Conventional injection moldingproducts surface defects are distinguished by artificial and efficiency and accuracy of thismethod are relatively low, which will bring harm to the health of the workers too. Machinevision technology, with a fully automated, high-speed, high-stability characteristic, is a verywide application in the field of defect detection. Injection molding products with polishedsurface have light reflective region on conditions of illumination, especially in the location ofcurved surface,which will make the process of image processing more difficult.Most criticalproblem is the reflective region may judge as a defect region segmentation，so quality of theproducts cannot be classified correctly. In addition, because the defect is very small and is notobvious compared with the color of the plastic products, threshold directly cannot segment it.In this paper, study on the image processing algorithms of common defects on thepolished surface of the injection molding products. Main work includes:1. Design machine vision systems hardware and software components. Focus on how toselect the camera, lens and the illumination program according to the actual detection.2. Introduce basic image processing methods of common defect detection and the use ofmixed programming of LabVIEW and Matlab. Propose a feedback automatic thresholdmethod using the number and shape of the connected region as a criterion. The effect of thismethod is better than a single method of automatic threshold when segment certain areas.3. Apply Homomorphic filtering to defect detection, using as image pretreatment.Combined with the region growing, mathematical morphology, edge detection method,segment some common defects of injection molding products successfully. The algorithm isnot sensitive to light,can solve the problem of reflection on polished surface of injectionmolding products, and be able to completely segment some defect whose gray value is similarwith its neighboring pixel.4. Use region description, mean gray value, moment invariants descript the defect region.Defects are classified by minimum distance classifier.