Research On Tobacco Foreign Material Online High-speed Pattern Recognition And Eliminate Technique

In recent years Material Online Nondestructive Examination (NDE) and Sorting System has been a point of issue in the research and application of industrial and agricultural automation. Tobacco Online Foreign Material Eliminating System (TFES), widely used in product line of tobacco factories to identify and eliminate foreign materials mixed in tobacco such as package cartoons, paper, moldy tobacco and other foreign materials, is an automatic photoelectric examination system. The success in the development of TFES is very significant to the improvement of the quality of tobacco products. The research into tobacco foreign materials online fast pattern recognition and eliminating system technique is the basis of the designing and development of TFES. This technique, which is ease to be put into application and can be applied in the fields of identifying foreign materials during the course of the production of other materials or eliminating from it, is very significant in practical production.This paper, based on the investigation into the present state of TFES development in both China and foreign countries from R&D of products and the research into the relevant key technique respectively, has made systematic and deep theoretic analyses and extensive and pertinent experimental researches to tobacco foreign materials online high-speed pattern recognition and eliminating system technique, has developed a set of tobacco foreign materials online fast pattern recognition technique with independent property right and has trial-produced the sample machine of TFES successfully according to this technique.The main contents of this paper are as follows: first, the statistical pattern recognition model is used to theoretically describe the problem of tobacco foreign materials recognition, in which some basic theories in the process of image segmentation, feature extraction, classification and decision making are analyzed. Aiming at the features that tobacco foreign matters pattern recognition requires high speed, several problems which is closely related to the realization of real-time in system are deeply discussed, such as threshold method in image segmentation, feature space and dimensionality reduction and minimum distance classifier. On the basis of these analyses, the Tobacco Foreign Materials Online High-Speed Pattern Recognition Model is worked out. And in this model, tobacco foreign materials pattern recognition problem was divided into the judgment problem existing in each judgment unit.According to all the data gained from the judgment units, we can decide the judgment units belong to tobacco or foreign materials.As for the real practice of some foreign materials are quite different from the tobacco in color and some other are adjacent to the tobacco, it was needed to distinguish foreign materials from tobacco by extracting effective features from two gradations of dot chrominance and judging units chrominance distributing. The content of detailed research includes: the principle and method of applying complementary colors to strengthen the gray difference between tobacco and foreign materials (based on the characteristic magnitude of grey); RGB color system and its extraction of feature and reduction of dimension, Ohta color system and HIS color system; the principle and method of identifying foreign materials from tobacco in the picture according to the evenness degree of grains and color distribution; the basic principle and method of extracting the chrominance difference of tobacco and foreign materials from the angle of changing fields. The research above indicates that many kinds of foreign materials could be effectively identified from tobacco by using the following three features-the hue of pixel, the first order central distance (or the standard deviation) of gray scale in discrimination unit.A dynamic threshold method-astute threshold method that is suitable for the condition of online identifcation is brought about, which would reduce the influence of the change of some conditions in th...