基于计算机视觉的岩屑图像识别与处理

Image Recognition and Processing of Rock Cuttings Based on Computer Vision

在传统的油田录井工作中,由于收集到的岩屑会受到地层的成分、结构和微观特征等诸多因素的干扰,且地层的岩性特征通常比较复杂,人为分辨岩屑种类存在主观性强、耗时长、误差大等问题,这都会在较大程度上降低岩屑描述的准确性。利用计算机语言中的OpenCV包,构建计算机视觉;依据HSV色彩模型,建立色彩跟踪器;通过调整HSV色彩值,可以识别不同岩屑的颜色,计算不同颜色的岩屑面积,实现对岩屑图像的分类与精细识别。利用数据挖掘中的K-means聚类算法,对岩屑数据进行分类,批量处理,得到了不同岩屑颜色的HSV色彩范围,完成了对岩屑图像的特征信息提取以及岩屑种类的划分与识别。通过对某一深度下采集出的岩屑样本数据进行测试集分析验证,将其结果与测井资料结论进行对比,确定其准确率达到93.4%。结果表明,利用基于K-means优化算法的岩屑识别分类可以对岩屑图像精确识别和处理,相对于人工识别分类而言,该方法在判断地层的岩性、油层位置、含油气性等信息中具有更高的准确性,可以为石油勘探开发提供更准确的地层信息。

In traditional oilfield logging work,the collected rock cuttings are affected by various factors such as the composition,structure,and microscopic characteristics of the formation due to various reasons.Moreover,the lithological characteristics of the strata are usually complex,and there are issues with strong subjectivity,long time consumption,and large errors in manually distinguishing the types of rock cuttings.This will greatly reduce the accuracy of rock cuttings description.The OpenCV package in computer language is used to build computer vision,and the color tracker is built according to the HSV color model.By adjusting the color HSV value,different rock cuttings colors can be identified,and the area of rock cuttings with different colors can be calculated to achieve classification and fine recognition of rock cuttings images.Using the K-means clustering algorithm in data mining,rock cuttings data was classified and processed in batches to obtain HSV color ranges for different rock cuttings colors.Completed the feature information extraction of rock cuttings images and the classification and recognition of rock cuttings types.By analyzing and verifying the test set of rock cuttings sample data collected at a certain depth,comparing the results with the conclusions of logging data,it is determined that the accuracy rate of this method reaches 93.4%.The results indicate that the rock cuttings recognition and classification based on K-means optimization algorithm can accurately recognize and process rock cuttings images.Compared to manual methods,this method has higher accuracy in determining information such as formation lithology,oil reservoir location,and oil and gas bearing properties.It can provide more accurate stratigraphic information for oil exploration and development.