Titanium nut surface defect detection technology is one of the key technologies to improve the competitiveness of enterprise products and improve the production process, while the traditional surface defect detection technology is difficult to meet the high-speed production needs, the implementation of machine vision-based titanium nut online inspection system to ensure the efficiency and accuracy of immediate inspection parts.
At present, although there are vision inspection systems used in industrial production, the machine vision-based online inspection technology for surface defects of titanium nuts is still in the research and development stage in China.
This project takes the high-speed inspection of titanium nuts as an example object, and studies the key algorithm of titanium nut defect detection based on machine vision, and deeply studies the titanium nut edge detection and area segmentation algorithm, and conducts experimental analysis of various algorithms.
The specific research content is divided into the following points.
1. image pre-processing: this paper firstly conducts a pre-processing study on the titanium nut images obtained online.
Pre-processing is an important pre-processing work for image processing and analysis, which directly affects the accuracy of image processing.
Image pre-processing includes two parts: filtering (denoising) and enhancement. In this paper, the classification and model of image noise are introduced, and the classical filtering methods are introduced in detail and analyzed experimentally, and then various image enhancement algorithms such as histogram equalization, Butterworth filtering and fuzzy theory-based image enhancement algorithms are used to process the titanium nut images respectively, and the experimental results are compared and analyzed.
2. Image segmentation: This paper is divided into two parts: image target segmentation and target region segmentation.
The image target segmentation aims to extract the whole titanium nut from the background, and the target region segmentation is to accurately segment the extracted titanium nut image to separate the bore, end face and gear parts.
In the process of image segmentation algorithm research, the image processing to be used are: pre-processing, titanium nut edge detection, titanium nut target extraction, titanium nut region segmentation, defect feature extraction, etc. Among them, pre-processing will use the first part of the experiment to derive part of the pre-processing algorithm and their combination, titanium nut target extraction segmentation using edge detection-based image segmentation technology, titanium nut region segmentation using a combination of fuzzy The final result is the complete extraction of different regional features of the titanium nut, which is the focus of this paper.
3. Defect detection: titanium nut surface defects including end defects, defects in the hole, gear defects three parts. In this paper, first of all, the detailed defect classification is carried out for different regions.
Then, combined with SVM theory, for each defect, multiple features and their fusion features are extracted and used in the SVM classifier to detect and identify the corresponding defects.
Finally, based on the experimental results, the detection rate of each feature is compared, and the optimal solution for titanium nut defect detection is proposed on this basis.
4. Test experiment: In order to apply the above algorithm theory to the titanium nut appearance defect detection system, this paper combined with the laboratory platform to conduct the corresponding test experiments to verify the effectiveness and real-time algorithm.