基于跨模态技术的地铁施工风险对策生成方法研究
Research on Intelligent Method for Generation of Subway Construction Risk Response Measures Based on Cross-modal Technology
周红 1周莉 1汤世隆 1黄文1
作者信息
- 1. 厦门大学 建筑与土木工程学院,福建 厦门 361000
- 折叠
摘要
为了给地铁施工现场人员应对风险提供智能辅助,提出了基于跨模态技术的地铁施工风险对策智能生成方法,通过采用改进的卷积神经网络ResNet50模型对施工现场风险图像进行语义特征提取,并利用LSTM模型和注意力机制融合图像和文本的语义特征,将施工要素的图像语义与文字语义相关联,以实现施工现场风险图像到风险对策的自动生成.经实验评价可知,提出的基于跨模态技术的地铁施工风险对策生成方法具有 0.9 以上的准确率和 0.8 以上的召回率.实现了根据采集的风险图像生成对策文本,为地铁施工风险应对阶段的智能辅助研究提供了可行有效的方法.
Abstract
To provide intelligent assistance for the subway construction site personnel to deal with risks,this paper proposed an intelligent generation method of subway construction risk response measures based on cross-modal fusion technology.An improved ResNet50 convolutional neural network model is used to extract semantic features from construction site risk images.The LSTM model and the attention mechanism are used to fuse the semantic features of the image and the text to predict the text corresponding to different parts of the image.Then,risk countermeasures could be automatically generated using construction site risk images.The experiment shows that the proposed subway construction risk countermeasure generation method based on cross-mode technology has an accuracy rate above 0.9 and a recall rate above 0.8,which has a good experimental effect.The research realizes the generation of countermeasure text according to the collected risk image,which provides a feasible and effective method for intelligent assistance in the risk response stage of subway construction.
关键词
地铁施工风险/跨模态生成/对策生成Key words
subway construction risk/cross-modal generation/response measures generation引用本文复制引用
基金项目
国家自然科学基金面上项目(71871192)
出版年
2024
NETL
NSTL
万方数据