The Analysis Of Quality And Determination Of The Geographic Origin Of Rice

The good quality of rice is effected not only by the genetic characteristics but also significantaffected by geographical environment and climate ecological environment.therefore, thequality of rice is obvious differences in different parts. In order to quickly identify thedifferent origin of rice, in this paper, the RGB images were obtained by the computer visiontechniques and near infrared grain analyzer was used to obtain the near infrered spectra of60kinds of brown rice and rice which from three origin. By image processing and near infraredspectra data analysis and BP neural network was finally used to identify origin of rice. Theresults showed that:(1) The median filter was used to remove the noise and and can keep the original information.(2) The different varieties of rice were recognized by BP neural network, and the structure ofBP neural network was7-16-11, the forecasting accuracy of feng Liang you was100%, fengyou si miao was92%, the lü huang zhan, hua you kang xiang nian and xin liang you821weremore than80%. But classification accuracy of tian you998, bai xiang139and hua you86was bad, only30%above.(3) Two kinds of methods were used to identify origin of rice by BP neural network: one wasusing the origin value not main component dimensional reduction, and the structure ofnetwork was19-42-2, the recognition accuracy of Guang Dong and Guang Xi were95%and88%respectively; the second was using the principal component to reduce the characteristicvalue, the network structure was7-16-2, the recognition accuracy of Guang Dong and GuangXi were95%and85%, respectively. There were not obvious differencesm but using theprincipal component dimension can reduce the time of every response, which can save time.(4) Standard normalization and scattering processing, and second order derivation, spectralintervals was4,smooth processing interval point was4, twice to smooth processing spectraprocessing was not used. Every2nm as a spectrum sections for extracting data, and principalcomponent analysis was used to reduce the spectrum data.(5)The structures of BP neural network of brown rice and rice was6-14-3and7-14-3,respectively. The detection accuracy of brown rice from Guang Xi province, Hu Bei provinceand Chong Qing was60%,80%and80%, respectively. And the accuracy of rice was80%,60%and100%, respectively.(6) Computer vision technology connect with near infrared spectra technology to identifythe origin of rice, there were little influence of hidden nodes on identifying the origin of rice.The accuracy of Chong Qing and Hu Bei were100%, and the accuracy of Guang Xi was80%.