Research Of The Technology Of Automated Cigarette Defect Detection Based On Machine Vision

Vision-based cigarette defect detection is one of the prevalent issues in the field of machine vision,image processing,pattern recognition,etc.During process of production and processing,maybe there are various defects may exist in the package of cigarettes when cigarettes are being transported from cigarette making and plug assembling machine to cigarette packing line.Common defects could be mainly summarized in three types:lack of cigarettes,contained foreign matter,and loose end(the end plane of cigarette is empty tobacco).To avoid aforementioned problems and in case the defective package flows into market,the defect detection in the process of package of cigarettes is adopted.In order to detect different kinds of cigarettes defects which consist of lack of cigarettes,foreign matter and loose end,this thesis presents a cigarette defects detection algorithm in the dynamic visual environment based on machine vision,image processing,and pattern recognition technologies.The proposed algorithm has important academic and application value.In this thesis,an approach of detecting cigarette packet defect such as lack of cigarettes,contained foreign matter,and loose end has been studied and proposed in the dynamic visual environment.At the same time,a cigarette defect detection system has been designed and implemented.At last,several experiments have been conducted to evaluate the performance of the detection.The mainly contents are as follows:(1)By analyzing cigarette image samples,a cigarette defects detection algorithm is selected,which focuses on the detection of cigarette arrangement,foreign matter and loose end.The method is capable of projecting the detection target feature.(2)For the key issue of cigarettes arranged region extraction and positioning,cigarette arrangement region extraction method based on projection of cigarettes and an adaptive positioning method based on model guidance are proposed.The experiment result illustrates the scope of application of the method.(3)The corresponding detection algorithm is designed for different cigarette defects detection.For cigarette foreign matter detection,a detection algorithm based on template matching is proposed.Meanwhile,the effect of the size of template on the foreign matter detection and template selection strategy are analyzed through experiment.For cigarette arrangement detection,an arrangement similarity index which is used to identify whether cigarette packet is lack of cigarettes and corresponding detection algorithm are proposed.Meanwhile,the selection strategy of arrangement detection threshold is analyzed through experiment.For cigarette loose end detection,a binarization method of cigarette loose end image based on iterative algorithm is proposed.On this basis,we propose a loose end similarity index for identifying whether cigarette is loose end and corresponding detection algorithm.Meanwhile,we show the selection strategy of loose end detection threshold through experiment.(4)According to the proposed cigarette defect detection algorithm a cigarette automatic defect detection system is designed and implemented.Meanwhile,the performance of the detection algorithm is evaluated by conducting experiments on the detection system.The experimental results prove that the detection algorithm meets the requirement of recognition rate and real-time in the process of cigarette detection.Experimental results indicate that the proposed algorithm can achieve the quality inspection and control in the cigarette production process by using machine vision,image processing and pattern recognition technologies.