Research On Key Techniques Of Vision Inspection For Injection Molding Products

In the past decades, along with the increasingly expanded application of plastic products, the renovation of the technics and equipments in plastic manufacturing process, especially in the injection molding (IM) process, has won broad concern, and considerable progress has also been achieved in this field. However, the development of the quality control technique which is the important tache of the IM process is still hanging behind and its automaticity is in low-level. Today, under the transformation of industry system from labor-density to technology-density, more and more deficiencies of current human-depending quality control manner have been exposed which will be the bottleneck of the IM process sooner or later. Therefore, we focus on computer vision based quality inspection of IM products in our work, in order to acquire accurate and comprehensive quality information for the realization of quality control automation. According to aforementioned considerations, several crucial technical problems have been considered in this thesis:(1) Information of production technology on IM process acquired by the approaches of on-site survey and literature reference is summarized and analyzed, and an experimental computer vision based inspection system of IM products is designed. Thereafter, the hardware and software platform is established and developed on the combination of process characteristics.(2) Preprocessing problems of IM product images that are easily contaminated by noise is studied, and a novel median filter--a region adaptive median filter which can not only filter the noise from product image but also protect the details of edges in it efficiently is proposed, which lays a solid foundation for the proceeding research of defect inspection.(3) Shape representation techniques of IM products are also considered in this work. Firstly, an improved morphological skeletonization algorithm is put forward in order to ensure the connectivity and medial axis property of the extracted skeleton. And then the problem of skeleton curve approximation is argued for the compression of the redundant information. An improved sequential polygonal approximation method is presented consequently, which effectively predigests useless trivial structure of the original skeleton while factually inherits its key structure. These novel strategies successfully pave the way for shape defect analyzing process, which includes shape representing, shape describing and shape reconstructing and so on.(4) Surface defect detection of IM products is another issue investigated in the thesis. A new adaptive structure element construction strategy as one of the crucial links of the following improved morphological edge detection algorithm is discussed firstly to promote the adaptability of the morphological operation. After that, the entire algorithm is described in detail. The noteworthy innovation is the combination of the morphological gradient and the differential direction which are indispensable for locating and extracting edge points from gradient images. Fortunately, all these improvements have ensured the acquisition of the accurate and precise edges of the IM products.(5) Feature extraction techniques of surface defects which are basilic for surface defects recognition and analysis are the last research task in this work. According to the characteristics, surface defects are divided into two primary classes. For the first class of defects which is known as isolated speckle-like and line-like defects, compound invariant moments that consist of Hu invariant moments and Zernike invariant moments are adopted as feature extraction scheme. On the other hand, an improved texture feature extraction algorithm based on Ridgelet transform is presented for the sake of the invariant feature extraction of the second class which is named texture-like defects owing to its property. Thereby, accurate and adequate feature information of surface defects is obtained for the further recognition and analysis.Finally, the potential further research direction in the area of quality inspection and quality control of IM process is discussed after summarizing the whole work in this thesis.