户外高光谱探测可以快速获取样品的光谱信息,但受环境光线和样品二向反射特性的影响,采集到的光谱并不能准确反映样品的真实信息,对户外探测精度有一定影响.为了提高户外高光谱探测精度,提出了一种使用空间特性光谱对户外光谱进行修正的方法,以冬枣、红提、小白杏为研究对象,使用Walthall、Shibayama、Ross-li、Roujean与Rahman这5种BRDF模型反演3种果品的空间特性光谱,利用反演的空间特性光谱对户外光谱进行修正,之后分别建立暗室光谱、户外光谱与修正光谱的品质预测模型.反演结果表明:3种果品的空间特性光谱均有较好的反演效果,反演误差从低到高依次为冬枣、小白杏、红提,平均决定系数R2分别为0.957、0.947、0.927,平均误差分别为3.56%、4.90%、8.23%;5种BRDF模型中,Walthall模型的反演效果最佳,平均决定系数R2与误差分别为0.949、5.33%,Ross-li模型的反演效果最差,平均决定系数与误差分别为0.934、6.05%.户外光谱修正结果表明:户外光谱经过修正后噪声减少,光谱更为平滑,且光谱趋势与暗室光谱一致,受反演效果影响,冬枣光谱的修正效果最佳,而红提与小白杏的修正光谱中噪声较多.品质预测模型结果表明:3种果品的品质预测模型预测效果差异较大,从高到低依次为冬枣、小白杏、红提,可能与3种果品的品质不同有关;5种BRDF模型得到的修正光谱所建立的模型预测效果不同,但无显著差异;预测模型中,使用暗室光谱建立的预测模型最优,修正光谱建立的模型预测能力优于户外光谱建立的模型,表明户外光谱经过修正后模型预测能力得到提升.BRDF模型能够较好地反演果品的空间特性光谱,修正后的光谱与暗室光谱较为接近,修正光谱建立的模型优于户外光谱建立的模型,表明使用空间特性光谱对户外光谱进行修正的方法是可行的,可为提高户外无损检测的精度提供一种新的思路.
Outdoor hyperspectral detection can quickly obtain the spectrum information of the sample,but affected by the ambient light and the bidirectional reflectance distribution function of the sample,the collected spectrum cannot accurately reflect the true information of the sample,which has a certain impact on the outdoor detection accuracy.In order to improve the accuracy of outdoor hyperspectral detection,a method of correcting outdoor spectrum using spatial characteristic spectrum was proposed.Walthall,Shibayama,Ross-Li,Roujean and Rahman were used to invert the spatial characteristic spectrum of winter jujube,red grapes and"Xiaobai apricot",the inverted spatial characteristic spectrum were used to correct the outdoor spectrum.The quality prediction models of darkroom spectrum,outdoor spectrum and correction spectrum were established respectively.The inversion results showed that the spatial characteristic spectrum of the three fruits has good inversion effects,and the inversion errors from low to high are winter jujube,"Xiaobai apricots"and red grapes.The average determination coefficient are 0.957,0.947,0.927,and the average errors are 3.56%,4.90%and 8.23%,respectively;among the five BRDF models,the Walthall model has the best inversion effect,the average determination coefficient and error are 0.949 and 5.33%,respectively.The Ross-li model has the worst inversion effect,the average determination coefficient and error are 0.934 and 6.050%,respectively.The outdoor spectrum correction results show that the noise of the outdoor spectrum is reduced after correction,the spectrum is smooth,and the spectrum trend is consistent with that of the darkroom spectrum;affected by the inversion accuracy,the correction effect of winter jujube spectrum is the best,and the spectrum is smoother,while the correction spectrum of red grapes and"Xiaobai apricot"have more noise.The results of the quality prediction model showed that the accuracy of the quality prediction model of three kinds of fruit was quite different,and the order from high to low are winter jujube,"Xiaobai apricot"and red grape,which might be related to the different quality of fruits;the prediction effects of the models established by the corrected spectrum obtained by the five BRDF models are different,but there is no significant difference;in the prediction model,the prediction model established by darkroom spectrum is the best,and the prediction ability of the model established by modified spectrum is better than that established by outdoor spectrum,indicating that the prediction ability of the model is improved after the outdoor spectrum is corrected.In summary,the BRDF model can better invert the spatial characteristic spectrum of fruits,the corrected spectrum is closer to the darkroom spectrum.The model established by the corrected spectrum is better than the model established by the outdoor spectrum,indicating that the method of using spatial characteristic spectrum to correct the outdoor spectrum is feasible and can provide a new idea for improving the accuracy of outdoor nondestructive testing.