| For being the key technology of part recognition, automatic identification technology is a very important study field. How to achieve effective marking and quick information recognition has long been a hot topic. DPM (Direct Part Marking) is a combination of two dimensional barcode technology and direct part processing, which takes use of some processing method and forms a permanent symbol on the surface of the part. The symbol labeled on the part then can be read by automatic reader, thus, information encoded is obtained whenever it is required. With high accuracy, stable performance and competitive rate laser marking is widely in part marking used to get legible and permanent label. It has extensive application prospect and high use value to study DPM on metal material for its high share in manufacture industry. However, its own texture, highly reflective behavior, poor resistance to rust and low-contrast all make it tricky to identify. A large amount of powerful readers have been developed by producers abroad, they can also provide machine vision solution, both intelligent camera and system based on PC included. In contrast, related products for machine vision system are usually labeled with poor accuracy, less reliability and poor university leading to the high dependence on import for good performance products. To expand the application of part recognition technology in manufacturing, it is urgent to improve technology to read direct part marks labeled on metal surface. Under this environment, this paper focuses on images of Data Matrix directly labeled on metal surface and the main contributions are listed below:Optimize the lighting condition of image collecting to improve the overall quality. Design an integrate image pre-processing solution, including image denoising, symbol locating and image binarization. Choose binarization based on iteration algorithm as the usual method. The result shows that the proposed algorithm gives better disposal effect and higher recognition rate is also gained. Compare documents relating to Data Matrix establishing or using standard, combining with features of the direct marks labeled on metal materials and determine quality factors used to quality evaluation. Contrast, print growth and axial uniformity are finally chosen. To construct the quality evaluating system, calculate the weight of each factor by CRITIC method and obtain the final grade of the symbol by TOPSIS method which is a multi-factors approach for quality assessment.Restore inferior images, low contrast and strong texture features included. Take the distribution of gray scale, edge information and local alteration of gray level into account and propose a novel method of image binarization, which is a combination of Otsu method and Nearest neighbor pixels contributing threshold method.Propose symbol reconstruction. Using the magnification feature of convolution operation improve the girding result of the symbol image. Make the first time reconstruction according to the gray level different between sub-regions and then conduct second reconstruction with the fixed structure of Data Matrix added. The comparison between the results of the two stage has shown that the necessary of the second time and the improvement of normalization and dependability of symbol reconstruction. |
PDF Full Text Request