The Studies On Quality Indexes And Evaluation Method Of Brown Rice

In order to research quality of brown rice and detection method, using Liao Xing as theexperiment material, this article researched influence on brown rice quality by unsoundkernels of brown rice, feature extracting and distinguish of brown rice kernels, method ofdetecting the milling yield, real milling yield and content of broken rice based on the digitalimage processing technology.(Ⅰ)The milling yield, real milling yield and content of broken rice increased, wholewell-milled rice ratio of brown rice and bulk density decreased with increase of unsoundkernels content. Water up-taking rate and expanded rate decrease, solids in cooking residualincreases, pH in cooking residual and starch-iodine blue value are disorder. The experimentshowed that three point bending force of sound kernels, chalky kernels immature kernels,dead kernels decreased respectively. The average of three point bending white and greenimmature kernels are only half of sound kernels’, these resulted in decrease of wholewell-milled rice ratio of brown rice and increase of broken kernels content. The average ofthree point bending of Long brown rice is higher than short brown rice, after milling, short isprone to broken rice.(Ⅱ)Brown rice was classified, classification as follow: sound kernels, chalky kernels,immature kernels, dead kernels, damaged kernels, broken kernels, yellow-colored kernels;After gray conversion, binaryzation, median filter, image segmentation, morphologicaloperations, noise and exceptional value were removed, image border became smooth,characteristic was extracted accurately. Feature of sound kernels, green sound kernels, chalkykernels (white, green), immature kernels (white, green), dead kernels (white, green), sproutedkernels, insect-damaged kernels, discolor kernels were extracted, described, and analysed,feature of sound kernels were different from other kernels generally, some difference wasobservable.(Ⅲ)After image pre-processing and extracting characteristic, the images information ofsound kernels, green sound kernels, chalky kernels (white, green), immature kernels (white,green), dead kernels (white, green), insect-damaged kernels, sprouted kernels, discolor kernels,were obtained. Carrying on the dimensionality reduction by the principal components analysisto the image information,7principal components were acquired, cumulative proportion in ANOVA was89.765%, the effective information was reserved. Establishing the neuralnetwork by the7principal components, training the neural network, recognizing unsoundkernels by the neural network, the recognizing accuracy of sound kernels, green sound kernels,chalky kernels (white, green), immature kernels (white, green), dead kernels (white, green),sprouted kernels, insect-damaged kernels, discolor kernels respectively was100%,96%,84%,88%,80%,84%,100%,90%,100%,86%,100%.(Ⅳ)After analyzing the correlation between the milling yield, the real milling yield andimage parameter, the correlation between the average of B (blue) and the milling yield,between H (hue) and the real milling yield was established. Detecting sample with theregression equation, the result showed that the result obtained the digital image processingmethod and the actual value was consistent. Using area as recognition parameter for headkernels and large broken kernels, large broken kernels and small kernels, the best thresholdwas obtained. The method of detecting head rice, large broken kernels, small broken kernelsby the digital image processing technology was established. The method of detecting headrice, large broken kernels, small broken kernels by the digital image processing technologywas established. Detecting sample with the regression equation, the result showed that theresult obtained the digital image processing method and the actual value was consistent.