Optimized Method Of Task Scheduling For Appearance Defects Automatic Detecting System Of Optical Filter

Filter as one of the key optical components in optical communication, its quality greatlyaffects the quality and the development of optical communication. Filter as an optical devicewith weak structure, manufacturing process is relatively complex, have a higher probability ofdefective goods. And filter the traditional detection technology is relatively backward, mainlyrely on the naked eye detection, filter quality is difficult to be guaranteed more unable to meetthe needs of enterprises of large-scale production, so to improve the filter appearancedetection efficiency has become the industry need to solve the problem. In this paper, thetopic of "scheduling optimization method" system task automatic detection of appearancedefects of optical communication filter, based on the working principle of the appearancedefect combined with information entropy feature selection filter detection designed filterappearance defect detection and classification of software, and to pursue several schedulingalgorithm in fact when respectively to study the software in the operation of the system canchoose, and the algorithm is optimized, which can realize the real-time fast and accurately tofilter the appearance of defects detection classification. The research results have importantacademic value and practical significance for promoting the development of filter detectiontheory and technology. The research work of this paper has been combining with the projectof Guangdong Province(Grant 2012B091100057) production.The paper firstly analyzes the overall demand filter defect detection, the design ofsoftware combined with information entropy feature selection filter appearance defectclassification based on appearance defects, and to study the classical scheduling algorithmfilter detection process and improve the scheduling algorithm based on filter appearancedefect detection based on the process, and to optimize the design, the experiments verify thefeasibility of the optimization method. The main work includes:The existing feature selection algorithm based on information theory, considering onlysingle correlation detection of appearance defects in current filter, a feature selectionalgorithm is presented based on joint entropy, its advantage is to explore the correlationbetween multiple defect characteristics and categories. Experimental analysis of thisalgorithm is derived, the classification performance than other algorithm and the algorithmhas the advantages of high, suitable for large sample, has quickly select the most relevantfeatures, 4 types of defects can correctly identify the filter, including point, marks, scratches,chipping, and the defect recognition correct rate is 100%;It makes the appearance defect filter detection process directed acyclic graph andimproved step by step, introduces two kinds of classical scheduling algorithm, task queuebefore defect detection process and first service based on short task priority task of detectingprocess based on depression;The appearance defect improved scheduling algorithm based on filter detection processtask, introduces two kinds of the improved scheduling algorithm based on task, optimizationof defect detection process time slice rotation method and based on the optimization task ofdefect detection process of multistage time slice method, and optimized the;The building appearance defect filter detection platform, CCD camer,alens, light sourceincludes an optical collection selection, design and operation of filter appearance defectclassification software and selected four different scheduling algorithms, finally theexperimental results are analyzed.The experimental results show that the appearance of defects classification, filtersoftware selection multistage time slice scheduling algorithm, photo finish time, defectdetection are the shortest completion time, 2/3 time saving than the typical model algorithm.