Development Of Optical Filters Surface Imperfections Vision Identification And Quality Rank Intellectual Evaluation System

Optical filters surface imperfections detection and quality rank evaluation depends largely on artificial. With the title "Development of Optical Filters Surface Imperfections Vision Identification and Quality Rank Intellectual Evaluation System", the thesis set up the surface imperfections detection and quality rank intelligent evaluation system. Achieve surface imperfections accurate real-time online detection and identification, implement quality intelligent rank evaluation to improve the efficiency of the filter surface detection. The research is supported by Guangdong province university-industry cooperation project(2012B091100057).The thesis first analyses the different among international optical elements criterion in the surface imperfections classification and quality rank evaluation method. Discusses advance of optical element surface i mperfections detection and assessment method, multiprocessor scheduling for image processing technology. Design filter surface imperfections detection and quality grade intelligent evaluation system, research on the filter surface visual identification se lf-learning optimization, multiprocessor scheduling optimization, to improve the accuracy and efficiency of imperfections detection. The surface quality of the filter level intelligence analysis algorithm, realize the filter surface quality rating. Set up the surface imperfections detection and quality rank intelligent evaluation system, tests are performed for detecting different filters. Main work of the thesis includes the following parts:? Implement the joint information entropy based on feature selection algorithm filter surface fault detection algorithm and double stages support vector machine classifier. Transplant the self-learning k-NN classification algorithm, which can pick the high confidence coefficient sample to expand the data set in the process of sample classification.? Analysis working-block of filter surface defect visual inspection system. Based on the topological relationship between the mission and draw AOE figure, Get the optimization parallelism test process AOE figure. According to the topological relationship after decoupling form a set of tasks. Raise the global APAs scheduling algorithm, based deadlines cache sensitive of tasks to a high level. Implementation of visual inspection system for multiprocessor scheduling, improve resource utilization.? Propose a tolerance-based surface quality rank analysis method, according to the method of Chinese national criterion, the surface quality can be divided into excellent products, superior product, subprime products, fine products, defective products. In this method, overall tolerance and inner effective aperture tolerance have different weight for the decision-making. A principal component analysis is taken for finding the principal component X of tolerance. A principal component X-based surface quality rank analysis method is proposed for eliminate the conversion process of two truncation errors. First truncation errors is surface imperfections points into the surface of the different fault tolerance series, another is tolerance s eries conversion to another tolerance series.? Meet the need of surface imperfection detection, set up the hardware of surface imperfections detection and quality rank intelligent evaluation system. According to the defect rate to select multiple sets of test samples, with the purpose of verifying the effectiveness of k-NN learning algorithm and multi-processor scheduling. Compared with workers rating results, proves the validity of surface imperfections recognition and quality rank evaluation.