准确监测土壤有机碳密度SOCD（Soil Organic Carbon Density）对调控土壤碳汇、合理利用土壤资源具有重要意义。机载高光谱影像为精细化SOCD制图提供了重要数据源。由于机载高光谱在数据收集过程中易受到外部因素的影响，光谱中存在噪声影响SOCD的估算精度。因此，本研究旨在探究基于机载高光谱影像估算SOCD的技术流程。对原始光谱进行预处理，包括一阶微分FD（First Derivative）和包络线去除CR（Continuum Removal）变换。采用遗传算法GA（Genetic Algorithm）选择特征波段，并结合不同回归方法，如偏最小二乘回归PLSR（Partial Least Square Regression）、多元线性回归MLR（Multiple Linear Regression）、支持向量机SVM（Support Vector Machine）和人工神经网络ANN（Artificial Neural Network）估算SOCD。结果表明，在经过GA特征波段选择后，原始光谱、FD光谱和CR光谱预测SOCD的精度均有所提高。使用原始光谱特征波段，PLSR、MLR、SVM和ANN四种模型预测SOCD的决定系数R2分别为0.672、0.721、0.551和0.678。使用FD与CR光谱特征波段的R2范围分别在0.452-0.593和0.332-0.602之间，具有较大的误差。利用原始光谱的特征波段进行SOCD数字制图，不同回归模型预测的SOCD在空间上具有较为相似的变化趋势，与SOCD测量值较为相近，绝对误差较大的点多出现在采样点边缘附近。

Accurate monitoring of soil organic carbon density (SOCD) is important for regulating soil carbon sinks and rational use of soil resources. Airborne hyperspectral images provide important data source for SOCD mapping. As airborne hyperspectral images are easily affected by external factors during data collection, the noise in the spectrum affects the accuracy of SOCD estimation. At present, a set of technical processes which is suitable for airborne hyperspectral data processing is still lacking. Therefore, this study aims to investigate the technical process of SOCD estimation based on airborne hyperspectral images. The original spectrum are pre-processed by first derivative (FD) and continuum removal (CR) transform. Genetic algorithm (GA) was used to select the feature bands. And different regression methods, such as partial least square regression (PLSR), multiple linear regression (MLR), support vector machine (SVM) and artificial neural network (ANN) were used to estimate SOCD. The results showed that the accuracy of SOCD prediction for original, FD and CR spectrum was improved after the feature bands.selected by GA. With the feature bands of original spectrum, the R2of SOCD predicted by PLSR, MLR, SVM and ANN are 0.672, 0.721, 0.551 and 0.678, respectively. The range of R2are 0.452-0.593 with FD feature bands, and 0.332-0.602 with CR feature bands, which have large errors. In this study, the feature bands of the original spectrum were used for SOCD mapping. The SOCD predicted by four regression models has a more similar trend in space and is similar to the SOCD measured value. The points with larger absolute errors mostly occurring near the edges of the sampling points.