论文部分内容阅读
针对隧道地质超前预报过程中,探地雷达(ground penetrating radar,GPR)线测图解释仅依靠专家经验,且存在准确率不高的问题,通过对GPR成像原理和隧道地质特性的研究,以及对深度置信网络(deep belief networks,DBN)计算复杂度的分析,提出一种改进的基于压缩感知和DBN的GPR线测图分类解释模型。该模型首先利用压缩感知技术对原始GPR线测图进行压缩处理,通过选择图像压缩比得到合理的压缩图像;然后将压缩后的图像送入DBN模型进行分类,根据分类结果对原始GPR线测图进行解释;最后利用广西六宜(六寨—宜州)高速公路隧道实测数据对模型的有效性进行验证,试验数据共20 000幅GPR图像,包括6种隧道地质类型,其中15 000幅图像作为训练样本集,5 000幅图像作为测试样本集。研究结果表明:当GPR线测图压缩比为256,反向微调数据为1 000幅图像,DBN模型迭代次数为30时,模型对测试数据中6类探地雷达线测图的分类准确率达100%,单次训练时间降低为原DBN模型的8%左右;大量仿真试验发现GPR线测图的合理图像压缩比区间为64~1 024,在此区间压缩的图像能最大限度地降低图像维度并且保留原始图像信息。该模型具有解释准确率高、训练速度快等优点,可为制定隧道施工和开挖计划提供合理依据。
In view of the advanced geological forecasting of the tunnel, the GPR interpretation only depends on the expert’s experience, and there is a problem that the accuracy is not high. Through the research of the GPR imaging principle and tunnel geological characteristics, This paper analyzes the computational complexity of deep belief networks (DBN), and proposes an improved classification classification model of GPR based on compressed sensing and DBN. The model firstly compresses the original GPR line by compressive sensing technology and obtains a reasonable compressed image by selecting the image compression ratio. The compressed image is then sent to the DBN model for classification. Based on the classification results, the original GPR line map Finally, the validity of the model is validated by the measured data from the tunnels of Liuyi-Yiizhou expressway in Guangxi. The test data includes a total of 20 000 GPR images, including 6 tunnel geological types, of which 15,000 images are used as Training sample sets, 5 000 images as a test sample set. The results show that when the compression ratio of GPR linemap is 256, the reversely tuned data is 1000 images and the number of iterations of DBN model is 30, the accuracy of the model in classifying the 6 GPR line charts in the test data is up to 100%, and the training time of single training is reduced to about 8% of the original DBN model. A large number of simulation experiments show that the reasonable image compression ratio range of the GPR linemap is 64-1 1024, and the image compressed in this interval can minimize the image dimension And retain the original image information. The model has the advantages of high accuracy of interpretation and fast training speed, which can provide a reasonable basis for the formulation of tunnel construction and excavation plans.