基于CT扫描和随机森林算法的沥青混合料油石比检测

为弥补路面施工过程中传统沥青混合料油石比检测方法耗时长、预测精度低、对环境及人身健康有危害等缺陷，利用CT扫描及图像处理技术得到的沥青混合料内部截面空隙与集料特征，提出了一种基于随机森林算法的沥青混合料油石比检测模型，并与多元线性回归算法、决策树算法、梯度提升树算法的预测结果进行对比。研究结果表明：CT扫描和图像处理技术对沥青混合料空隙和集料的识别效果良好；通过比较分析不同算法的平均绝对误差、均方根误差、平均绝对百分比误差和负相关系数，证明采用随机森林算法的沥青混合料油石比检测模型的准确性能够满足工程应用要求，且优于其他3类算法模型，相比于传统检测方法更加简便、快捷。

Asphalt-Aggregate Ratio Acquisition for Asphalt Mixture Based on CT Scanning and Random Forest Algorithm

Traditional acquisition method of asphalt-aggregate ratio during the construction of asphalt pavement is time consuming, with low accuracy and harmfulness to the environment and human health. Alternatively, this study tries to obtain the characteristics of void and aggregates through cutting-edge CT scanning and image processing technologies, while based on which an asphalt-aggregate-ratio acquisition model is further established based on random forest algorithm. The test results are compared with those obtained from methods of multiple linear regression, decision tree and gradient boosted decision tree. The results show that a combined use of CT scanning and image processing technology could well predict volume contents of voids and aggregates within asphalt mixture. By comparing MAE, RMSE, MAPE and R2 of four different algorithms, it is proved that the accuracy of the asphalt ratio test model of asphalt mixture based on the random forest algorithm can meet the requirements of engineering application, which is simpler and faster than the traditional test method.