The Dimensional Inspection And Surface Defect Recognition Of Injection Molded Products Based On Machine Vision

With the development of injection molding technology, the dimensional accuracy and surface quality of injection molded products have become demanding increasingly. However, the dimensional inspection and surface defect recognition of injection molded products mainly rely on manual sampling detection today, which is low degree of automation and inefficient. The machine vision detection technique has characteristics of non-contact, high flexible and efficient, so it can achieve high efficiency, automation and intelligent detection tasks instead of artificial detection technology. This paper aims to use machine vision technology to inspect the key dimensions and recognize the surface defects of the injection molded products.This paper designs a system that used to the dimensional inspection and surface defect recognition of the injection molded products, designs the overall structures of hardware and software system, studies on the selection of the key hardware such as light source and camera, and designs the software modules. The paper develops the graphic user interface with MFC as well.The image processing technology is the core technique of machine vision systems. This paper studies the key algorithms such as the image gray, image enhancement, image filtering, image segmentation and image edge detection algorithms which used in the dimensional inspection system. In addition, the paper chooses the appropriate algorithms on the basis of experiments to get edges of the product, which lays a good foundation for the further dimension measurement.Based on the study of the Hough Transform, this paper proposes a method which can extract the line feature of the product image appropriately. When using the method to recognize the line feature of the image, the paper finds that the method can decrease the memory space used and computing time comparing to use the Hough Transform instead. On the basis of extracting the line feature of the image successfully, the paper realizes the measurement of the key dimensions of the product. In addition, the paper verifies the accuracy of the measurement on the basis of experiments, and analyses the error causes as well.An intelligent recognition method based on the convolution neural network is proposed to find the short shot and weld lines defects by focus on the study of the short shot and weld lines recognition of injection molded products. The method can effectively solve the problem that existing recognition algorithms need to extract features artificially, use heuristic methods and professional knowledge. Based on the structure of the common convolution neural network, the method modifies a lot and optimizes the parameters of the network. Experiment indicates that the recognition accuracy rate for short shot, weld lines is 99.91% and 99.08%.