Research On Plastic Product Defects Detection Algorithm Based On Machine Vision

Plastic, as a closely linked polymer materials with our life, is increasingly used invarious fields like construction, industry and agriculture. Substantial growth in the productionof plastic products puts forward higher request to the quality control in the process of theproduction. And the detection of the plastic product defects is one of the key links in theplastic production process to improve production efficiency and quality-control precision,which has great practical significance and value.The subject comes from the education combined project of Guangdong ProvinceMinistry named High Efficient Visual Detection Technology and Equipment of ElectronicProduct Appearance Surface Defects (Project Number2012B091000014). Based on themachine vision we make research on the detection and recognition of surface defects onplastic products. We study the related theory and technology on the defects detection ofplastic products. The image processing and defect recognition algorithms are mainlydiscussed. The main research contents of this article are as follows:First, introduce the development status of machine vision defects of plastic products,leading to vision detection technology work process. Point out the three difficulties ofmachine vision algorithm and explain the topic origin and the research goals. Based on themachine vision theory, introduce the system hardware selection and then build the systemtesting platform.Second, from the view of image processing algorithms, do research on plastic products,contrast different results and select the most suitable algorithm. Use a variety of filteringalgorithm on plastic product image to delete noises in the photos and find the median filter isthe most efficient one. In edge detection, study many algorithms and select the mostappropriate LOG algorithm as the best one.Finally, focus on the detection algorithms. For tilt image Hough transform could make it.Template matching algorithm is used to target detection products and locate interferencecharacteristic regions. Compared with the traditional mask, the partial filling algorithm isproved to be more effective to fill the interference characteristic regions. After analyzing the shortcomings of traditional LOG algorithm, propose a modified LOG detection algorithm tosolve the problem of texture defects detection. Experiments show that the defect detectionalgorithm in this paper detects the main defects on plastic products like scratches、holes,which has certain application value.