초록 ▼

○ 연구결과
우유 시료에 Methyl Red, Methylene Blue, Bromcresol Purple, Phenol red, Resazurin, Eosin-Y, Amido Black의 반응 시약을 첨가한 후, 고성능 분광광도계(NIRS 6500)를 이용하여 원유의 실시간 휴대용 품질판정 시스템 개발을 위한 전처리 조건 및 적정 파장 영역을 분석한 결과는 다음과 같다.
- 우유 시료의 온도는 유성분의 온도에 따른 영향을 최소화하기 위해 40 ℃로 유지하였다.
- 체세포수 측정을 위한 환경조건은 Resazurin

○ 연구결과
우유 시료에 Methyl Red, Methylene Blue, Bromcresol Purple, Phenol red, Resazurin, Eosin-Y, Amido Black의 반응 시약을 첨가한 후, 고성능 분광광도계(NIRS 6500)를 이용하여 원유의 실시간 휴대용 품질판정 시스템 개발을 위한 전처리 조건 및 적정 파장 영역을 분석한 결과는 다음과 같다.
- 우유 시료의 온도는 유성분의 온도에 따른 영향을 최소화하기 위해 40 ℃로 유지하였다.
- 체세포수 측정을 위한 환경조건은 Resazurin 시약을 사용하여 반응 시간을 5분으로 사용한 경우가 400～600 nm 영역에서 가장 좋은 상관관계를 보였다.
- 지방, 단백질, 유당, 총고형분 측정을 위한 환경조건은 반응 시약의 첨가 유․무 및 반응 시간과 상관없이 1,400～1,600 nm 영역대에서 가장 좋은 상관관계를 보였다.
- 그러므로 원유의 실시간 휴대용 품질판정 시스템 개발을 위한 전처리 조건은 체세포수와 지방, 단백질, 유당, 총고형분의 동시 측정을 위해 Resazurin 시약과 5분간 반응하는 것을 전처리 조건으로 설정하였고 적정 파장 영역은 체세포수의 경우 가시광선 영역으로, 지방, 단백질, 유당, 총고형분의 유성분은 근적외선 영역으로 선정하였다.
고성능 분광광도계 를 이용하여 (NIRS 6500) 전처리 조건 및 적정 파장 영역을 분석한 결과를 기본으로 전처리 장치(시료공급 장치, 시약투여 장치, 온도제어 장치, 시료혼합 장치), 분광장치(광원부, 검출부, 광섬유, 샘플홀더), 우유의 스펙트럼 데이터 수집을 위한 인터페이스 모듈과 전체 시스템 제어를 위한 고성능 임베디드 장치를 포함한 원유의 휴대용 실시간 품질판정 시스템을 개발하였다. 또한 실시간 품질판정을 위해 시스템 구동 프로그램과 품질판정 알고리즘 및 분석 프로그램을 개발하여 원유의 휴대용 실시간 품질판정 시스템의 예측 성능을 평가한 결과는 다음과 같다.
- 체세포수는 400～700 nm에서 상관관계를 보였으며, 검증한 결과 상관계수는 0.85, 오차는 13,111 개로 나타났다.
- 지방은 1,200～1,500 nm에서 상관관계를 보였으며, 검증한 결과 상관계수는 0.95, 오차는 0.45 %로 나타났다.
- 단백질은 1,400～1,600 nm에서 상관관계를 보였으며, 검증한 결과 상관계수는 0.92, 오차는 0.23 %로 나타났다.
- 유당은 1,300～1,600 nm에서 상관관계를 보였으며, 검증한 결과 상관계수는 0.86, 오차는 0.33 %로 나타났다.
- 총고형분은 1,300～1,500 nm에서 상관관계를 보였으며, 검증한 결과 상관계수는 0.92, 오차는 0.38 %로 나타났다.
- 원유의 휴대용 실시간 품질판정 시스템의 원유 품질요인 측정에 대한 정확도는 상용화된 우유 품질분석장치에 비해 다소 낮은 예측력을 보여주고 있으나, 현장에서 실시간 품질판정을 위한 휴대성과 고가의 우유 품질분석장치 보다 경제성이 우수하여 축산 농가 및 낙농 현장에서 유용하게 사용될 것으로 판단된다.

Abstract ▼

Because the quality of milk has influence on the cost of milk products, measurement of milk properties such as SCC, fat, protein, SNF and lactose is the most important for dairy industry. Especially, the SCC is a recognized indicator of cow's health and milk quality. Milk from infected cows, associa

Because the quality of milk has influence on the cost of milk products, measurement of milk properties such as SCC, fat, protein, SNF and lactose is the most important for dairy industry. Especially, the SCC is a recognized indicator of cow's health and milk quality. Milk from infected cows, associated with mastitis, has the high level SCC. Fast detection of SCC can help to exclude infected milk from dairy production and early treatment can minimize losses of dairy farmers. To improve the quality of milk, frequent measurement of milk composition of every individual cow was needed. The industries use several types of sophisticated instruments to determine milk constituents in the laboratory. SCC is determined by and integrated flow cytometer. Fat, protein, SNF and lactose are determined by infrared spectroscopy in the range of 3,500-9,600 nm. But reference method was expensive and time-consuming and there are currently no practical on-farm methods to determine milk properties. The NIR spectroscopic technique, which has very high signal to noise ratio, has a very short detector response time, requires minimum to no sample preparation, can simultaneously determine multiple milk properties in the farm. The final goal of this research is to develop a real-time portable quality evaluation system which can determine milk properties simultaneously in the dairy farm. The real-time portable quality evaluation system consisted of a flow pump to supply milk sample and reaction reagent at given volume, a fan to mix milk with reaction reagent, a thermoelectric device to maintain sample temperatures of ${40^{\circ}C}$, tungsten-halogen lamp, spectrometer, optical fiber probe to measure milk spectrum, and embedded module to control signals. The performance of the real-time portable quality evaluation system was analyzed by comparing the prediction accuracy of the laboratory pectrophotometer. Total of 200 milk samples were collected from dairy farms in the central areas of Korea. The milk samples were preserved to minimize propagation of bacteria cells during transportation. Milk properties such as SCC, fat, protein, SNF and lactose, were measured by the Combifoss (Foss-Electric, Model 5000, Denmark) and used as reference values. Reagents such as Methyl red, Methylene blue and Resazurin for reduction test, Eosin-Y for dying cell and Amido black for dying protein were added to milk samples. They activated for 5 minutes as optimal reaction time (KFRI, 2006) before spectra measurements. The transmittance spectra were measured in wavelength ranges of 350∼1,700 nm. five repetitive scans were averaged, transformed to log(1/Reflectance), and then were stored in the embedded module, forming one spectrum. Spectra of milk samples were divided into a calibration set and a validation set. Samples were ranked by values of SCC, and each set was selected by increasing rank evenly. Half of the total samples were selected for calibration set, and 50% were reserved for validation set. The calibration set was used during model development, and the validation set was used to predict milk properties from unknown spectra. The method of PLS analysis was used to determine the milk properties. A unique set of PLS loading vectors (factors) was developed. The multiplicative scatter correction(MSC) and the first derivative pretreatments based on Savitzki-Golay polynomial filter were applied to spectra to minimize sample-to-sample light scatter differences. A developed spectrum analysis was used to perform the PLS analyses. For each constituent, up to 20 factors was examined. Cross validation was performed during model development, where one of the calibration samples at a time was removed from the calibration set. The standard error of calibration (SEC) was considered to determine the optimal number of factors and the range of wavelength during calibration. On completion of the calibration, the model was used to predict milk properties from the validation set. Model performance was reported as the correlation coefficient (r),the standard error of prediction (SEP), and the average difference between measured and predicted values (bias). The PLS models showed good correlations between predicted and measured SCC with reagents in visible ranges. The PLS analyses for fat, protein, SNF and lactose showed good relationship with reagents in the NIR ranges. The PLS models showed excellent correlation to predict SCC by adding Resazurin, fat, protein and SNF. These PLS models might be good enough to predict SCC, fat, protein and SNF by applying spectroscopic techniques in the diary farm. The correlation coefficient of lactose was lower than the other constituent models, but the PLS models showed the ossibility to determine the variance of lactose. The prediction performance can be enhanced when the noises of reflectance signals are reduced and method for signal processing is improved.