Research And Implementation Of Liquid Dangerous Goods Classification Algorithm Based On Low-rank And Sparse Decomposition

In recent years,as the trend of globalization becomes more and more obvious,the contact between countries in the global scope becomes more and more frequent.All kinds of vehicles are carriers with frequent mobility of people and high concentration of high-value goods,so the safety of public places has been highly concerned by people.At present,X-ray detector is the most common security device in public places in China.Security inspection work has been greatly improved,and the incidence of accidents has been reduced at the same time.However,X-ray detector is not suitable for the security check of liquid dangerous goods.Therefore,if a safe and efficient detection method of liquid dangerous goods can be proposed,the safety of public places can be ensured,people's travel will be more convenient,and the social harmony and stability can be further maintained.On this basis,this paper takes the study of low-rank sparse decomposition algorithm in the classification direction of liquid dangerous goods as the subject,and from the perspective of signal classification,provides a new idea and method for the accurate classification of liquid dangerous goods.After reading a lot of literature,this paper investigates and studies the relevant algorithms for the classification of liquid dangerous goods.and the low-rank and sparse decomposition algorithm is analyzed.In this paper,from the perspective of liquid signal processing,a beam focusing system is designed to collect the S parameters of liquid dangerous goods samples.By extracting the S parameters of different types of liquids,a sample database is established.Then,low-rank and sparse decomposition algorithm is adopted to classify different types of liquids.The specific research results are as follows:1?In this experiment,the signal acquisition system combined with the vector network analyzer is used to collect the ultra-wideband centimeter-wave signals of three common security liquids in the range of 8ghz-18 ghz frequency band.In order to facilitate the evaluation of the algorithm,the collected centimeter-wave signals are set up.Due to the problems of frequency band loss,sample duplication,amplitude data and phase data confusion in the data set itself,the cleaning operation was adopted.Meanwhile,the data had problems of poor identity and a large amount of noise,so the operation of noise reduction and filling was adopted.2?This paper proposes a method for background liquid signal extraction based on low-rank sparse decomposition.Through the study of a large number of documents and the analysis of the actual situation of signal acquisition,it is concluded that the liquid signal is composed of the superposition of low-rank background liquid signal,sparse abnormal task signal and noise signal.First,the data set is constructed in the form of a data matrix,and the Go Dec algorithm is used to decompose the liquid detection signal data into a low rank part,a sparse part,and a noise part.Since most of the sparse matrix is abnormal data and filling data,it cannot be represented as a stable signal feature,so the background liquid signal part of low rank will be used for subsequent algorithm processing.3?Feature extraction of background liquid signal.The feature extraction of signals is the process of obtaining information features from signals.In this paper,a signal time-frequency analysis method based on empirical wavelet transform is adopted.In this algorithm,the background liquid signals are divided adaptively,the component signals are extracted,and then the time-frequency diagram of the signal is obtained by Hilbert transformation of each component signal.After analyzing the time and frequency of the signal,the eigenvectors of the marginal spectrum of empirical wavelet are extracted and used as the basis of classification.4?Training and recognition of sample signals.Because of the applicability of BP neural network to the detection signal of liquid dangerous goods,BP neural network is used to classify and identify the detection signal of liquid dangerous goods.Finally,the low-rank sparse decomposition algorithm is compared with the depth algorithm.This paper makes a comparison from the aspects of running speed and accuracy,and finds that the low-rank sparse decomposition algorithm has a better classification effect.The accuracy of the low-rank sparse decomposition algorithm for the classification of liquid dangerous goods can reach 92.6%,and the classification accuracy is greatly improved,which provides a new idea for the classification of liquid dangerous goods in the future.