基于高光谱成像技术无损检测芒果轻微损伤

Non-destructive detection of minor damage in mangoes based on hyperspectral imaging technology

中文关键词:  高光谱成像技术  芒果  轻微损伤  无损检测

英文关键词:hyperspectral imaging technology  mango  minor damage  non-destructive detection

中文摘要:

      目的 使用高光谱成像技术实现对芒果轻微损伤的无损识别。方法 在可见光-近红外波长范围内采集完好芒果和损伤芒果的高光谱图像, 并提取相应的感兴趣区域(regions of interest, ROI)获得样本高光谱数据。经过多种预处理方法比较, 选择光谱预处理方法。使用竞争性自适应重加权算法(competitive adaptive reweighted sampling, CARS)和连续投影算法(successive projections algorithm, SPA)分别对预处理后的光谱提取特征波长, 并分别建立了多元线型回归(multiple linear regression, MLR)模型和偏最小二乘回归(partial least squares regression, PLSR)模型。结果 选择多元散射校正(multiplicative scatter correction, MSC)作为光谱预处理方法。针对芒果轻微损伤识别, CARS-MLR模型识别效果最好, 其校正集相关系数为0.881, 预测集相关系数为0.821, 校正集均方根误差(calibration set root mean square error, RMSEC)为0.146, 预测集均方根误差(prediction set root mean square error, RMSEP)为0.236, 准确率为97.14%。结论 利用高光谱成像技术可以实现对芒果表面轻微损伤进行有效鉴别。

英文摘要:

      Objective To realize the non-destructive identification of minor damage of mangoes by hyperspectral imaging technology. Methods The hyperspectral images of intact and damaged mangoes were collected in the range of visible and near infrared wavelengths, and the corresponding regions of interest (ROI) were extracted to obtain the hyperspectral data of the samples. The spectral preprocessing method by comparing various preprocessing methods. The competitive adaptive reweighted sampling (CARS) and successive projections algorithm (SPA) were used to extract characteristic wavelengths from the preprocessed spectra, respectively. With the extracted characteristic wavelengths, a multiple linear regression (MLR) model and a partial least squares regression (PLSR) model were established, respectively. Results The multiplicative scatter correction (MSC) was selected as the spectral preprocessing method. The CARS-MLR model had the best performance in identifying mangoes minor damage, calibration set correlation coefficient was 0.881, prediction set correlation coefficient was 0.821, calibration set root mean square error (RMSEC) was 0.146, prediction set root mean square error (RMSEP) was 0.236, and the accuracy was 97.14%. Conclusion Hyperspectral imaging technique can be used to identify minor damage of mangoes surface effectively.