Classification Of Rice Origin In Heilongjiang Province Based On Cross-media Feature Fusion

As a large rice-growing province,Heilongjiang Province has an annual output of rice at the forefront of the country.However,due to the different quality of rice in different production areas of Heilongjiang Province,the rice in different production areas is also different.This has led to some businesses performing sub-optimal actions to obtain illegal profits,resulting in the destruction and loss of information on the origin of rice.Its reference value when people buy rice.Therefore,in order to solve the current chaos problem in the rice industry,a safe origin tracing method for the rice origin is proposed.Rice safety traceability technology includes near infrared spectroscopy technology,computer vision detection technology,etc.Due to the long detection period of chemical traceability technology,it is difficult to achieve real-time online detection,while physical traceability technology has a faster detection speed but lower accuracy,which reduces the reliability of practical applications.Therefore,this paper proposes a cross-media feature fusion based The traceability method of joint detection of near infrared spectroscopy and computer vision aims to solve the problem of accuracy and reliability of rice real-time online traceability of origin.In this paper,there are three methods of rice origin classification in Heilongjiang Province: image-based research on rice origin classification in Heilongjiang,near-infrared spectroscopy based rice origin classification in Heilongjiang,and cross-media feature fusion in Heilongjiang province rice origin classification.In the research of rice origin classification based on image processing in Heilongjiang Province,the image was de-noised,image binarization and image feature extraction were carried out respectively,and combined with different algorithms to carry out rice origin classification experiment in Heilongjiang Province.In the research on the classification of rice origin in Heilongjiang Province based on near infrared spectroscopy,the pretreatment and feature extraction of near infrared spectroscopy were carried out,and on the basis of this,different classification algorithms were used to carry out the experiment of rice origin classification in Heilongjiang Province.Finally,based on the extraction of rice image features and rice near-infrared spectral features,the Cartesian product is used to map the two views of rice from low-dimensional to high-dimensional shared space,and second-order fitting is used to fit the coupling relationship between features.And based on support vector machine,BP neural network and random forest three machine learning algorithms and cross-media feature fusion modeling,the rice origin of Heilongjiang Province was classified.The comparison of the rice origin classification of cross-media feature fusion with the rice origin classification of image features and near infrared spectroscopy.The experimental results show that the cross-media feature fusion method has a higher classification accuracy than the single rice origin classification method.Therefore,the comprehensive analysis method of cross-media fusion using the data features obtained by near infrared spectroscopy and the appearance of rice images can not only perform real-time online detection of the rice origin,but also avoid the use of a single near infrared spectroscopy analysis method for experimental results Produce inaccuracy and unreliability.