基于光谱特性分析的冬枣渐变损伤研究

Study on the Development of Bruises on Winter Jujube Based on Spectrum Characteristics Analysis

为了较早地对冬枣损伤进行预测,减少冬枣内部损伤引起的储藏损失,以山东沾化冬枣为研究对象,对冬枣内部隐性损伤直至表面微观损伤的渐变光谱特性进行研究,利用高光谱成像系统采集每个冬枣在同一试验条件下的各个损伤时期的高光谱图像,得到波长在390~1090 nm的512幅高光谱分量图像,从表面微观损伤的感兴趣区域反推内部隐性损伤的感兴趣区域,并获取各个损伤时期的光谱信息,通过3组差谱分析并交叉验证,确定变化较大的8个波长,再根据冬枣内部的主要成分变化确定4个波长,最终选取528.5、547.4、573.9、702.7、727.2、755.7、880.2、942.6、982.7、518.5、863.0、973.4 nm 12个波长作为冬枣渐变损伤的特征波长。利用偏最小二乘分析方法建立判别模型,并对预测集的83个样本(无损伤27、第1阶段损伤12、第2阶段损伤12、第3阶段损伤12、第4阶段损伤10、第5阶段损伤10)进行预测,检测精度依次为100%、58.3%、66.7%、83.3%、100%、100%,总体检测精度为86.7%。

In order to predict the damage of winter jujube earlier and reduce the storage loss,Shandong Zhanhua winter jujube were taken as the research object to study the gradual spectral characteristics of the hidden damage from winter jujube to the microscopic damage on the surface.Hyperspectral images of each period of each jujube under the same test conditions at various damage periods were collected to obtain 512 hyperspectral component images with wavelengths in the range of 390-1090 nm,and internal hidden damage was inferred from the surface micro-damage area of interest and the spectral information of each injury period were obtained.Through three sets of differential spectrum analysis and cross-validation,8 wavelengths with large changes were determined,and then 4 wavelengths were determined according to the main component changes in winter jujube.Finally,528.5,547.4,573.9,702.7,727.2,755.7,880.2,942.6,982.7,518.5,863.0,973.4 nm 12 wavelengths were selected as the characteristic wavelength of the gradual damage of winter jujube.A partial least squares analysis method was used to establish a discriminant model,and 83 samples of the prediction set(no damage 27,first stage damage 12,second stage damage 12,third stage damage 12,fourth stage damage 10,fifth Stage damage 10)for prediction,the detection accuracy was 100%,58.3%,66.7%,83.3%,100%,100%,and the overall detection accuracy was 86.7%.