基于机器视觉的印刷质量在线检测技术研究-印刷包装技术与设备专业论文.docxVIP

AB ABSTRACT 西安理 西安理工大学博士学位论文 i万方数据 i 万方数据 ii万方数据 ii 万方数据 Title:RESEARCH ON THE ON-LINE DETECTION TECHNOLOGY OF PRINTING QUALITY BASED ON MACHINE VISION Major：Printing Packaging Technology and Equipment Name：Jian-ming Sun Signature: Supervisor: Prof. Shi-sheng Zhou Signature: ABSTRACT With the rapid development of todays society digitalization and informatization, the mode of printing quality control manually has been unable to adapt to the actual needs of the printing production. Especially in the process of various invoice printing, the number character is the most important identification feature for products using variable code printing method so that the disordered number would cause huge loss. Therefore, in order to completely guarantee the printing quality and efficiency, it is very important to carry out online quality detection and identification for printing images and characters. By means of data collection, field experiments and analysis related to changeable printing information, defect types, color difference and so on, this thesis studies the online printing quality detection techniques based on the computer vision, and develops an online quality detection system to perform a complete online quality detection. The main achievements in this work are as follows: Aiming to the image segmentation methods for invoice with complex shading background, this thesis firstly uses the morphology-based multi-scale edge detection algorithm to extract the edges of the invoice image, and then segments the images based on the threshold adaptively determined by the marginal image gradient histogram. Thus, the shading background is effectively removed and better segmentation is achieved. This thesis comes up with a character image feature extraction method combining the circumferential projection transform, wavelet analysis and DCT algorithms. Experimental results show that this method possesses the properties of scale and rotation invariance, and good anti-noise c