研究目的
To evaluate hyperspectral imaging as a technique to predict the total nitrogen concentration of almond kernels.
研究成果
Hyperspectral imaging demonstrated great potential for non-destructive prediction of total nitrogen concentration in almond kernels, with a PLSR model achieving R2 of 0.82. This technique can be used for rapid assessment along the almond supply chain, including at retail outlets, to ensure food quality.
研究不足
The study used a limited number of almond brands (four) and samples (100 images), which may not represent all variations. The model's performance (R2 of 0.82) indicates potential but could be improved with more data or refined methods. External factors like kernel size or storage conditions were not fully controlled.
1：Experimental Design and Method Selection:
The study used a hyperspectral imaging system in the 400-1000 nm range to capture images of almond kernels. A partial least squares regression (PLSR) model was developed to correlate spectral data with laboratory-measured total nitrogen concentration. Data pre-processing included Savitzky-Golay 1st order derivative transformation and leave-one-out cross-validation.
2：Sample Selection and Data Sources:
Four brands of almond kernels were purchased from retail supermarkets in Australia. Each sample consisted of two whole kernels, with 100 images captured (25 per brand).
3：List of Experimental Equipment and Materials:
Hyperspectral imaging system (Pika XC2 camera, SpectrononPro software), LECO TruSpec analyser for nitrogen measurement, black background, white teflon board for calibration, and almond kernels.
4：Experimental Procedures and Operational Workflow:
Kernels were imaged in a dark room using the hyperspectral system with specific exposure time and scanning speed. Images were calibrated using dark and white references. Spectral data were extracted and processed. Total nitrogen concentration was measured destructively using combustion analysis.
5：Data Analysis Methods:
Spectral data were analyzed using Unscrambler X software. PLSR model was applied with performance metrics such as R2, RMSE, and RPD evaluated.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
