A5003804640- Drilling and Well Engineering
- Tunneling and Rock Mechanics
- Image Processing and 3D Reconstruction
- Mineral Processing and Grinding
- Fault Detection and Control Systems
- Soil Geostatistics and Mapping
- Advanced machining processes and optimization
- Structural Health Monitoring Techniques
- Image and Signal Denoising Methods
- Engineering and Test Systems
- Advanced Neural Network Applications
- Advanced Data Compression Techniques
- Infrastructure Maintenance and Monitoring
- Remote Sensing and LiDAR Applications
- Advanced Image Processing Techniques
- Engineering Diagnostics and Reliability
Beijing Jiaotong University
2023-2024
摘要: 土体含水率是影响细粒土性质的主要因素.土体表层含水率的快速识别是农业和岩土工程中智能监测和智能建造技术发展中的急迫需求.为了克服传统含水率测量或监测方法无法满足土体表层含水率的实时无损监测的局限性,特研发基于图像的含水率智能识别算法.首先在实验室中收集了4种不同类别的土体、在不同含水率下的表面照片,获得了超过1 400张图片的高质量样本库,为机器学习模型构建奠定了数据基础.然后采用经典的卷积神经网络对土体含水率图像数据集进行学习,建立了土体含水率智能识别模型.模型比选结果表明:基于ResNet34架构的土体含水率识别模型效果最佳,在测试集上的含水率预测平均误差约为2%.该模型初步满足了土体表层含水率的实时无损监测需求,能够为农业和岩土工程中智能监测和智能建造技术发展提供重要手段. 关键词: 土体含水率 / 深度学习 卷积神经网络 智能监测 智能建造 工程地质
Coarse soil is frequently utilized in filling projects, such as road foundations and earth-rock dams, its gradation affects the quality of filling. With rapid advancement intelligent construction techniques, there a growing need for fast recognition gradation. The conventional method acquiring coarse soil, sieving, time-consuming reliant on manual labor. A model based images has been proposed to address this issue. database 22080 photos was collected, first time, problem limited samples...