Research And Implementation Of Key Algorithm For Appearance Defect Detection Of Titanium Nut

The detection technique for titanium nut's surface defect is one of the key technologies to improve the enterprise competitiveness of products and production process while the traditional inspection technology of surface defect is difficult to meet the demand of high-speed production. The implementation of online detection system of titanium nut based on machine vision makes it possible to produce the parts online with high speed and high precision. At present, although the automatic detection technology has been a certain application in industrial production, the online detection technology of titanium nut's surface defects based on machine vision is still in the period of research and development in China. In this paper, the high-speed detection of titanium nut is taken as an example to study the defect detection algorithm based on machine vision. The algorithms of edge detection and region segmentation are studied deeply while a variety of algorithms are tested and analyzed. Specific research contents of this paper are divided into the following points.1. Image preprocessing: In this paper, the pretreatment of the titanium nut image is studied firstly. Preprocessing is an important work in image processing and analysis,which directly affects the precision of image processing. Image preprocessing includes two parts: filtering (denoising) and enhancement. This paper introduces the classification and model of image noise while made a detailed introduction and experimental analysis for classical filtering methods. Then a variety of image enhancement algorithms are used in image processing of titanium nut, such as histogram equalization, Butterworth filter, fuzzy theory and so on. At last the experimental results were compared and analyzed.2. Image segmentation: It is divided into two parts in the research of image segmentation: image segmentation for extracting the overall goal and target region segmentation. Among them, the first part is designed to extract the whole target from the background while the second part is aimed at the region segmentation of titanium nut which include the inner hole, the end face and the gear parts. Many algorithms will be used in image segmentation such as preprocessing, edge detection, target extraction,region segmentation, feature extraction and so on. The pretreatment algorithms and the combination of them in Chapter 1 will be used for preprocessing. There is a variety of detection operators will be used in edge detection for experimental comparison.Threshold segmentation based on edge detection is used for target extraction. Seed region growing method based on fuzzy theory is used to extract the feature of different regions which is the focus of this paper.3.Defect detection: the surface defects of titanium nut include three parts: the end defect, the hole defect and the gear defect. In this paper, we first classify the defects in different regions. Then, combined with the SVM theory, we extract various features of each defect and apply it to the SVM classifier respectively. Finally, according to the results of the experiment, the detection rate of each feature is compared, and then the optimal scheme for the defect detection of titanium nut is put forward.4. Test experiment: In order to apply the above algorithm theory to the titanium nut appearance defect detection system. In the present paper, we carry out the corresponding test experiment based on the laboratory platform to verify the effectiveness and real-time performance of the algorithm.