基于高光谱成像的肥城桃品质可视化分析与成熟度检测

Visual Detection of SSC and Firmness and Maturity Prediction for Feicheng Peach by Using Hyperspectral Imaging

肥城桃采摘后转色快、易腐烂,导致果品等级下降。采用高光谱成像技术对其进行可溶性固形物含量(SSC)和硬度可视化分析与成熟度检测,以提高果品质量,实现优果优价。首先,采集成熟度为70%和90%的各80个肥城桃的高光谱信息、SSC和硬度,通过蒙特卡罗偏最小二乘法分析剔除异常值,利用光谱理化值共生距离划分样本集,采用竞争性自适应权重采样算法(CARS)和连续投影算法(SPA)选取特征波长,并建立多元线性回归(MLR)模型。研究表明:CARS-MLR模型性能优于SPA-MLR模型;预测SSC的CARS-MLR模型,Rc^2和Rv^2分别为0.8191和0.8439,RPD为2.0;预测硬度的CARS-MLR模型,Rc^2和Rv^2分别为0.9518和0.8772,RPD为2.1。然后,基于CARS-MLR模型计算肥城桃每个像素点的SSC和硬度,生成可视化分布图,实现不同成熟度肥城桃SSC和硬度可视化检测。最后,利用顺序前向选择算法优选特征波长,建立人工神经网络成熟度预测模型,获得98.3%总识别准确率。

Feicheng peach is prone to spoilage due to its surface color changing rapidly after harvest,which will degrade its quality.Hyperspectral imaging technology was used to detect the soluble solid content(SSC),firmness and maturity of Feicheng peach for improving its quality and price.There were 80 maturity 70%and 90%Feicheng peach were used for hyperspectral images(400~1000 nm),SSC and firmness collection,respectively.These samples were split into calibration set and validation set with a ratio of 2∶1 by samples set partitioning based on joint X-Y distances method after the outliers were eliminated by using Monte Carlo-partial least squares method.MLR detection models were established using feature wavelengths selected by competitive adaptive reweighted sampling(CARS)and successive projections algorithm(SPA),respectively.The more effective detection results was emerged by CARS-MLR model,with a determination coefficient of calibration set(Rc^2)of 0.8191,a determination coefficient of validation set(Rv^2)of 0.8439 and a residual prediction deviation(RPD)of 2.0 for SSC assessment,Rc^2 of 0.9518,Rv^2 of 0.8772 and RPD of 2.1 for firmness assessment.Visualization maps for SSC and firmness were generated by calculating the spectral response of each pixel on peach samples.Furthermore,the artificial neural network model was provided to predict the maturity of Feicheng peach using feature wavelengths selected by the sequential forward selection algorithm,with total recognition accuracy of 98.3%.It can be concluded that hyperspectral imaging technology can be applied to determine the SSC,firmness and maturity of Feicheng peach,laying a foundation for the on-line nondestructive quality monitoring and timely harvest of Feicheng peach.