比较分析空间分辨光谱和单点可见/近红外光谱(可见/短波近红外光谱和中波近红外光谱)对番茄颜色的识别能力。 根据番茄表面和内部颜色将600个样品分为6个等级(green, breaker, turning, pink, light red和red)。 分别利用新型空间分辨光谱系统(550~1 650 nm), 可见/短波近红外光谱仪(400~1 100 nm)和中波近红外光谱仪(900~1 700 nm)采集番茄的空间分辨(spatially-resolved, SR)光谱和单点可见/近红外(SP Vis/NIR)光谱, 建立番茄等级的偏最小二乘判别(PLSDA)模型, 比较其对番茄颜色等级的预测效果。 结果表明, SR光谱组合可在最佳单一SR光谱基础上进一步提高番茄颜色的识别能力, 对番茄表面颜色和内部颜色的识别率可分别达到98.8%和84.6%。 光源-检测器距离较近的SR光谱对番茄表面颜色的识别有帮助, 而光源-检测器距离较远的SR光谱能较好的判别番茄内部颜色。 SP NIR光谱在对番茄表面颜色判别中与SR光谱具有一定可比性, 其分类准确率可达到95%, 但SP Vis/NIR光谱在对番茄内部颜色识别中具有较低的分类准确率, 分类结果远不如SR光谱, 说明SR光谱比SP Vis/NIR光谱对番茄颜色的判别更具潜力。

The paper reported the comparison of recognition for tomato surface color and internal color by spatially resolved and conventional single point visible and near infrared (SP Vis/NIR) spectroscopy. Spatially resolved (SR) spectra and SP Vis/NIR spectra were acquired using the newly spatially resolved spectroscopy system (wavelength: 550~1 650 nm), the portable Vis/NIR spectrometer (wavelength: 400~1 100 nm) and the portable NIR spectrometer (wavelength: 900~1 700 nm), for 600 “Sun Bright” tomatoes with six color stages (green, breaker, turning, pink, light red and red), based on their surface and internal color distribution, respectively. Partial least squares discriminant analysis (PLSDA) models for SR spectra and SP Vis/NIR spectra were developed and compared. The results showed combination of the SR spectra could further improve the classification of tomato color based on optimal single SR spectra, with classification accuracy for surface and internal color of 98.8% and 84.6%, respectively. The SR spectra with short source-detector distance were useful for recognition of tomato surface color, while SR spectra with large source-detector distance could better assess tomato internal color. The NIR spectra were comparable with SR spectra for tomato surface recognition with classification accuracy of 95%, however, SP Vis/NIR spectra could not evaluate tomato internal color accurately, and the classification accuracy was much lower than that of SR spectra, which indicated that SR spectra have great potential for the recognition of tomato color.