Visual Inspection Methods And Technology For Advanced Electronics Manufacturing Production Line

Advanced electronic manufacturing is a fundamental and strategic industry for national economy development. As a core technology in industrial automation and intelligent, advanced electronics manufacturing vision automatic detection and recognition is a sunrise industry with broad prospects, and it carries a huge market potential filled with unlimited business opportunities. As an emerging area of technology, machine vision detection and recognition in the international arena is still in the initial stages of development, and it is still in the land reclamation seeding in our country. Study of advanced machine vision detection theory and methods and advanced technology is very essential for improving China's electronics manufacturing automation and intelligence level.This dissertation first introduces the related processes and products in electronics manufacturing, considers domestic and international Machine Vision Detection of the status quo to extract the scientific and technical issues to be solved. Subsequently, this dissertation introduces the key hardware and software systems of advanced electronics manufacturing.Focused on the visual inspection problems in advanced electronic manufacturing and combined with the National Natural Science Foundation of China-"High-speed precision manufacturing production line optimization of visual inspection and intelligent control technology research", this dissertation has studied advanced electronic manufacturing method of visual detection theory and key technology research, and it mainly including the following aspects of the work.(1) A image denoising method is proposed for visual inspection of the advanced electronics manufacturing. After a brief introduction of conventional image denoising method, this part presents a four tree complex wavelet packet transform approach based on a mixed statistical model of image noise suppression new method. In the proposed method, the noisy image is decomposed into a low-frequency approximation sub-images and a number of high-frequency direction of sub-graph, only the direction of sub-maps of high-frequency noise suppression. The use of complex correlation coefficient between layers the direction of the strength of the high-frequency sub-images are divided into major categories and secondary categories, respectively, and the corresponding statistical model approach has achieved better results.(2) In order to study the quality PCB machine vision-based intelligent detection algorithms, this part presents a neural network based intelligence PCB quality detection algorithm, and a parallel chaos optimization algorithm based PCB component detection algorithm. In the PCB quality control, this method selects the appropriate characteristic parameters as neural network input according to the normal and failure of solder joints of different image features, thus the establishment of neural network classifiers can detect PCB solder joint quality. In the PCB component detection algorithm, it first selects the appropriate template, as templates and search sub-graph matching to identify the component type of parallel chaotic optimization algorithm to search the best matching so as to identify the component.(3) An intelligent mounter visual positioning method is presented for the advanced electronic manufacturing line. In this method, first wavelet transform is used to extract the image characteristics of electronic components, and then RBF neural network pattern matching algorithm is used to achieve accurate positioning placement machine vision.(4) A chip-pin edge detection method is proposed based on multi-structuring elements and multi-scale morphological. First, the part presents a morphological edge detection operator, using the detection operator for image edge extraction, and then re-uses of multi-structuring elements morphological structure of multi-scale edge information elements adjustments. In addition, study of a chip based on Hough transform to extract the appearance parameter method, using Hough transform has been extracted from the edge pixel information to seek the specific appearance of the chip parameters.(5) For electronic materials (magnetic ring) quality testing, a magnetic ring quality intelligent detection method is proposed based on gray-scale histogram and support vector machine. This part uses low-dimensional gray scale information to describe the characteristics of magnetic ring, and the image after separation from the background to carry out processing to extract the gray histogram feature. Then it uses principal component analysis, statistical information will be gray by the high-dimensional vector reduced to low-dimensional vector. Subsequently, a low-dimensional vector as input, using support vector machines for classification in order to achieve the quality of intelligence testing magnetic ring.At last, a major innovative research result of this research work is concluded and it carried out on the next prospect.