Detection Technology Based On Machine Vision In The Plastic Injection Molding Process

Injection molding is one of the most common forming methods of plastic products. In the plastic injection molding process, products will appear kinds of defects. Now the surface qualities of products mainly rely on manual sampling detection, which is of low automation degree and efficiency. It is a common phenomenon that products ejection incomplete, which has bad effects on products quality and the mold. Machine vision is a new technology which can take place of manual detection, providing accurate and efficient automation solutions for the detection task. This paper aims to use machine vision technology to realize the products online surface defects detection and automatic identification, as well as the detection of cavity abnormal situation.This paper analyzes the composition and principle of machine vision system and designs a injection molding process detection system based on machine vision combining with the actual situation, for online surface defects automatic detection and identification, as well as automatic detection of cavity abnormal situation. Based on these, this paper designs the overall structures of hardware and software system, studies on the key hardware selection, and designs the software processes and main software modules, and develops the graphics user interface with Qt.To reduce the interference of the non-target area and improve the detection accuracy, the paper sets ROI manually, and supports many ROIS; according to the software process of products surface defects detection, this paper firstly studies the key algorithms such as the image enhancement, template matching, image segmentation, morphology processing, contour extraction, and chooses the appropriate algorithms on the basis of experiment to get areas of surfaces defects; based on principles of these algorithms, this paper realizes automatic detection of cavity abnormal situation by using image filtering, template matching algorithms, and experiments prove the accuracy of detection.In order to overcome the shortcomings of traditional defects recognition such as the BP neural network and SVM support vector machine (SVM) algorithm, which all need training product images samples and bad adaptability, experiments are finished to obtain the data of image features in view of a short shot, flash, crack these three defects, and selects some image features that are useful for defects recognition, then this paper puts forward a defects classification algorithm for these three defects, experiments prove this defects identification algorithm is feasible.The software system developed in this paper basically achieves automatic and intelligent detections of products surface defects and cavity abnormal situation in the process of injection molding, which is of high practical value.