Digital image analysis of outcropping sediments

The possibility to use colour data, a fast and inexpensive method of proxy data generation, extracted from two selected loess-paleosol sequences is discussed here. We compare the outcome from analysing outcrop images taking by digital cameras in the field and spectral colour data as determined under controlled laboratory conditions. By nature, differences can be expected due to differences in illumination, moisture, and sample preparation. Outcrop inclination may be an issue for photographs; correcting for this is possible when marks can be used for rectification. In both cases the data extracted from images match the visual impression of photos well, and are useful for obtaining a more quantitative measure for field observations. Smoothness (as measured by autocorrelation) is high for an image from Achenheim/France, where an image with a width of ca. 1.1 m and a depth of 1.6 m was analysed. Data from a narrower image part from Sanovita/Romania are noisier. In both example cases, a significant correlation between data extracted by digital image analysis and laboratory measurements could be established, suggesting that image analysis may be a useful tool where outcrop- and light-conditions allow useful photographs, especially where high resolution proxy data is required.本文探讨了利用从两组精选黄土-古土壤序列（loess-paleosol sequences）中提取的颜色数据的可行性——该方法是一种快速且低成本的代理数据（proxy data）生成手段。我们对比了两类分析结果：一类是对野外数码相机拍摄的露头图像（outcrop images）开展分析所得的结果，另一类是在受控实验室条件下测得的光谱颜色数据的分析结果。本质而言，由于光照、湿度与样品制备方式的差异，两类数据存在差异在所难免。露头倾角可能会对拍摄的照片造成干扰，但若可借助标记物进行校正，则可修正该偏差。两种场景下，从图像中提取的数据均与照片的视觉观感高度契合，且可为野外观测提供更为量化的评估维度。来自法国阿兴海姆（Achenheim）的图像提取数据平滑度（以自相关（autocorrelation）系数衡量）较高，该分析图像覆盖了约宽1.1米、深1.6米的露头区域。来自罗马尼亚萨诺维塔（Sanovita）的较窄图像区域提取的数据则噪声更强。在上述两个示例案例中，数字图像分析提取的数据与实验室测量数据均呈现出显著相关性，这表明，在露头与光照条件可获取优质照片的场景下，尤其是在需要高分辨率代理数据的研究中，图像分析可成为一种实用的研究工具。