| Summary: | 碩士 === 國立雲林科技大學 === 資訊工程系 === 105 === In recent years, digital multimedia application knowledge is flourishing and a large number of life, business and industry, in order to cope with the iron and steel industry, the follow-up needs of steel products, this paper to small steel embryo depth image and its automatic identification of small steel embryo Application as the main research direction.
In this paper, the research method of industrial camera and "single-point direct-type" laser scanning, you can get a small steel embryo digital laser scanning images, the calculation can be obtained after the small steel embryo digital laser scanning images Dimensional information, and proposed to establish a simple and effective small steel embryo depth image recognition model, and proposed to extend the local binary rotation invariant mode to detect small steel embryo depth image recognition model texture features, and to histogram homogenization Histogram Equalization was used to identify the texture distribution of the small steel embryo depth image recognition model. Finally, we wanted to identify the texture distribution information of 100 samples and 100 sorting samples in the database, sort and output the matching order according to Euclidean Distance of the identity of the results, the sort of the front is expressed as the higher the similarity of the sample.
In this paper, the depth of the small steel embryo image and the establishment of automatic small steel embryo identification, in this experiment to Euclidean distance Euclidean Distance comparison and statistical analysis of the identification results, the first order, the first five The correct rate for each of the top ten identifiers is 100%, 95% and 85%, respectively.
|
|---|
