中国西南丘陵常态山和喀斯特山交错分布，遥感影像普遍存在山体阴影，分布零散且无规律，基于DEM的地形校正模型（C校正等）虽然算法成熟、易于操作，但在复杂地形区存在误差。引入基于相似像元包络线的阴影校正方法（CR校正），按照阴影提取、包络线去除、相似像元寻找和阴影亮度重建的步骤，采用西南丘陵地区Landsat 8 OLI影像进行验证实验。结果表明：CR校正后，阴影区的视觉特征与邻近非阴影区趋于一致，阴影像元亮度有明显提升；校正后影像主要波段标准差减小，与非阴影区参考光谱的相对均方根误差在2.919%以内，最低仅为0.516%；自动分类精度从43.59%提高到61.57%，CR校正有效提高了有阴影的丘陵地区遥感影像质量。

Shadow of the remote sensing image is widespread in China southwest hilly area,which has affected the effects of automatic recognition of image computer and quantitative analysis.Topographic correction methods,which are widely used to adjust for differences in solar incidence angles,can partly alleviate the impacts of shadows.But the model based on DEM data has limitations and errors in application,resulting to the scattered and discontinuous images of the terrain correction.In order to overcome the shortcomings of the topographic correction models based on DEM and improve the quality of remote sensing image of southwest hills,the article introduced a new shadow correction methods based on the similar spectral information after continuum removed.The advantage of the method is that it no longer depend on DEM and semi-empirical estimation value,which can maximize the computer automatic identification and the independent access for parameters.
Throughout these shadow correction methods,they are based on weak information in the shadow area,and the shadow correction is achieved by establishing a relationship between shadow and non-shadow.As to shadow pixel and non-shadow pixel with the same surface land cover types,their spectral curve is similar and only brightness (continuum information) is different.In order to establish relationship of spectral statistic feature between shadow pixel and non-shadow pixel,the spectral information is recovered by the similar pixel of shadow pixel and by the principle of using surface features spectrum envelope line to remove the continuum line.The article introduced a new shadow correction methods based on the similar spectral information after continuum removed,and experiment it on Landsat 8 OLI image by shadow extraction,continuum removing,searching for similar pixel,shadow information restoration.And the calibration accuracy was tested by visual evaluation,statistical analysis,comparative verification and automatic classification.
By comparing the images of C-method correction and CR-method correction,the visual feature of the two images tend to be flat,and the image details of shadow area tend to be obvious.After CR-method shadow correction,the pixel brightness value of shadow area gets a more complete compensation,and the image brightness converges with the shadow area,and it makes the shadow area visual characteristics more consistent with the near shadow area.At the same time,the standard deviation of the image tend to drop after CR-method correction, making the pixel brightness value of the slopes closer.CR-method is better than C-method for terrain shadow elimination and slope and luminance values of homogenization,which relative root mean square error (rRMSE) of sample point in per land use cover type is within 2.919% compared with unshaded pixels,and the minimum is only 0.516%.the automatic classification accuracy of CR-method correction, calculating the number of right and wrong pixels,has increased from 43.59% to 61.57%.The experiment of CR-method shadow correction in complex hilly terrain region has achieved good effect,and has improved the quality of remote sensing image with mountain shadow.