| Summary: | 碩士 === 國立臺灣科技大學 === 纖維及高分子研究所 === 86 === With the elevation of industrial technology and life horizon, manufacturing industries are replacing human workers with automatic artifices gradually for benefits like higher efficiency and lower costs. Recently, the production of synthetic fiber in Taiwan expands vigorously especially that of polyester texture yarn, which has become number 1 in the world in terms of quantity. Most of the false twist texturing factory have been highly automated, but all of them encounter a bottleneck of complete automation, for inspecting the appearance of yarn packages still requires human labors to examine each product one by one. This task is not only non-objective but also greatly heighten up the cost of labor in the process. Moreover, the wage in Taiwan has been higher and higher today, so that modification of this process has become a problem that is eagerly needing to be solved. Hence we designed an inspection station specially for the appearance of false twist yarn packages, combining techniques of machine vision and digital control. To finish this task in an automatic manner is expected. Some skills of image processing including convolution filtering, image segmentation, curve fitting, signature extraction and morphological operation, are implemented with programming language. The size of a package, discoloration, formation and overthrow at the bottom of package are inspected. We integrated the algorithms with two CCDs, rotating plate driven by a servo motor and illumination equipment with a personal computer, following the concepts of visual inspection system. Thus a practicable, low-cost and automatic visual inspection system was exploited in this study. This system can turn the indeterminate concepts into explicit representation by extracting index for each characteristic of the package. These results were used to grade the products using theories of statistical pattern recognition. After the accomplishment of the construction, several samples were examined to test the functions and effects. According to the data obtained, the result diagnosed by the system meet the fact, providing a standard for classification. Comparing to traditional method of human labor, it is more precise and performs a more regular classification.
|
|---|
