环境一号卫星CCD相机(HJ-1-CCD) 30m的空间分辨率在地物识别中具有潜在优势， 然而由于缺少短波红外通道， 利用暗像元法反演地表反射率较为困难。 基于北京与珠三角地区的地物光谱试验， 获得暗像元的植被指数与红、 蓝波段反射率比值， 构建基于辐射传输模型的大气校正算法。 为了验证算法精度， 将北京地区卫星反演值与实测的草坪、 水体、 沥青、 水泥等光谱数据， 以及居民区、 建筑物等模拟数据进行对比分析， 表明红、 蓝波段反演的相对误差分别控制在38.7%和37.2%以内。 通过与MODIS地表反射率标准产品对比， 当反射率较小时红、 蓝波段的相关性R2分别达到0.809 4和0.723 9； 环境星CCD数据能有效减弱混合像元影响， 对水泥、 建筑物等亮目标的反演精度高于MODIS产品。

The potentiality of the retrieval of surface reflectance using CCD camera aboard HJ-1A/B satellite was studied. It is very difficult to use dark targets in atmospheric correction due to the lack of near infrared band. The alternative normalized difference vegetation index (NDVI) and the red/blue reflectance ratio are detected from the spectral experiment in Beijing and the Pearl River Delta. Ground-based spectral data including grass, dense vegetation, water body, soil, residential roof and bright building etc. were used to validate the surface reflectance in Beijing, and the relative error in red, blue band is under 38.7% and 37.2% respectively. Uncertainties of the surface reflectance retrievals were analyzed. The comparison of MODIS surface reflectance product showed that there is a good agreement in the dense targets, and the correlation coefficient (R2) in red, blue band is as high as 0.809 4 and 0.723 9 respectively. HJ-1-CCD data can effectively reduce pixel-mixed impact on the cement roof and bright buildings, and the inversion accuracy is higher than MODIS products.