http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
노상하 서울대학교 농업개발연구소 1999 농업생명과학연구 Vol.3 No.-
Sugar and acid in fruit are essencial chemical components which are highly concerned to fruit taste and quality. An on-line sorting system to measure the sugar and acid contents in fruit was developed. The sensing device of the system was composed of a high sensitivity CCD spectrometer, a fiber optic probe and tungsten-hallogen light source so that the transmittance spectrum of the fruit sample could be measured. The fruit samples were fed into the sensing device by an automatic feeder and conveying equipment. PLS regression models to predict the sugar content and acid content were developed with 210 Fuji apples for calibration and another 210 for prediction. Main factors affecting the prediction error were radiation intensity of light source, and conveying speed and temperature of the apple. SEP and R2 of the sugar prediction model were 0.525 Brix% and 0.8, and those of acid prediction were 0.05% and 0.29, respectively, at the sorting speed of 3 apples per second. Based on these results it was concluded that with the on-line sorter Fuji apples could be classified into three grades by sugar content and two grade by acid content.
라인스켄 카메라를 이용한 현미의 온라인 품위판정 시스템 개발
노상하 서울대학교 농업생명과학대학 농업개발연구소 1998 농업생명과학연구 Vol.2 No.-
The aim of this study was to develop an automatic brown rice inspection system which could discriminate the brown rice kernels into the sound, cracked, chalky, green transparent, green opaque, white opaque, colored and red kernels. Primarily, an automatic feeding device was designed with a vibrator, flat belt conveyor and parallel rails to align and individualize the brown rice kernels in bulk state. Secondarily, discrimination algorithm was developed with a line-scan camera system equipped with optical fiber illumination. It was found that important factors to discriminate the brown rice kernels were length and cross section area of kernels, gray value gradient in lengthwise and histogram patterns of the kernel images. Discrimination accuracies of the neuro-net based algorithm using those factors were ranged from 85% to 97% except for the green-transparent and the red kernel samples. However, performance of the whole system including the feeding device was not satisfactory for on-site use. Further study on illumination method is required to improve the performance.