一种基于随机森林与颜色特征的岩屑识别算法

在随钻过程中，如何自动、快速、客观地识别岩屑是录井过程中一个十分重要的问题.针对这个问题，对岩屑颜色进行了深入的分析研究，对不同岩性的岩屑图像提取颜色直方图，将其作为岩屑特征，最后将随机森林法应用到岩屑的识别过程中，取得了较好的识别效果.通过对某气田几口井的现场随钻，结果证明本文算法识别正确率达到90％以上，为快速自动录井提供了一条有效的途径.

A cutting recognition algorithm based on randon forest and color

In the cutting logging process, how to recognize the cuttings automatically and correctly is very important. In this paper, the features of cuttings are detected by using Color Histogram. Then the features are used to train the randon forest classifier. Finally, the cuttings are classified by using the randon forest. Using the algorithm, after logged several gas reservoir, a high recognition rate can be reached at 90%. The experimental results demonstrate the algorithm is feasible, robust and applicable.