Research And Application Of Cigarette Appearance Quality Detection Algorithm

Surface defect is an important and visible factor that affects the appearance quality of cigarette.It mainly refers to macular,puncture,pinch,fold and other defects different from cigarette thread paper.In recent years,most of the market complaints received by tobacco enterprises are caused by the appearance quality of cigarettes.Therefore,the detection of cigarette appearance quality is of great significance.At present,the detection of cigarette appearance quality in cigarette factories is mainly through manual sampling and visual classification.This method has some problems,such as low detection efficiency,difficult to unify the judgment standard,easy to cause secondary pollution of cigarettes and so on.In view of these problems,this paper makes a more in-depth study on the detection methods of cigarette appearance quality.From the perspective of machine learning,an image feature extraction algorithm based on multi feature fusion is proposed,and the classification detection of cigarette image is realized by using support vector machine(SVM);from the perspective of deep learning,a convolutional neural network model suitable for cigarette appearance quality detection is studied and applied in practice.This study will provide a new idea and method for the detection of cigarette appearance quality.The specific research contents and work are as follows:Firstly,the preparation process of cigarette experimental samples is introduced,and the cigarette image acquisition,preprocessing and image positioning are studied,including image denoising algorithm and cigarette image positioning algorithm based on FFT template matching.Secondly,from the perspective of machine learning,we study a variety of cigarette image feature extraction algorithms,such as the traditional local binary pattern(LBP)algorithm,the improved equivalent pattern(ulbp)algorithm,and the directional gradient histogram(HOG)algorithm.Aiming at the problem that a single feature can not well describe the image feature information,we propose a cigarette image feature extraction algorithm based on multi feature fusion In this algorithm,the local texture features extracted by ulbp and the edge texture features extracted by hog are fused to form a new feature matrix.Finally,SVM classification model is used to realize image classification detection.Then,from the perspective of deep learning,the cigarette appearance quality detection algorithm based on convolution neural network is studied.This paper introduces the development background of deep learning and the basic principle of convolutional neural network,constructs the convolutional neural network model by using the GPU version of the python deep learning framework,and determines the optimal performance neural network classification model through a series of super parameter comparative experiments.This method has a new breakthrough in time complexity and classification accuracy compared with the traditional machine learning method.Finally,in order to realize the practical application of cigarette appearance quality detection algorithm,the overall design of the system is carried out from two aspects of hardware and software,and the system design principle and device selection are described,and the performance of the system is verified.From the test results,the cigarette appearance quality detection system can meet the actual detection needs of cigarette factory.