基于人工智能的清水墙典型损伤智能识别与评估方法研究

Research on Intelligent Identification and Evaluation Method of Typical Damage of Fair Faced Wall Based on Artificial Intelligence

为避免对清水墙典型损伤进行查勘时,传统人工查勘方法存在人为因素干扰大、手动标注效率低、目测识别精度低的问题,研究了基于人工智能的清水墙典型损伤智能识别与评估方法。采用深度学习与计算机视觉方法,搭建了基于YOLO V4的单阶段目标检测神经网络模型,建立了风化、泛碱、绿植覆盖3种清水墙典型损伤的深度学习神经网络模型训练图像数据库,实现了上述3种清水墙典型损伤的智能诊断及区域划分。同时,基于OpenCV计算机视觉库,分别计算了像素尺寸级别下的各类损伤区域面积比,以代表清水墙损伤的程度。工程实践证明,与传统人工查勘方法相比,基于人工智能的方法实现了高效、准确、便捷的清水墙损伤自动识别与快速评估流程,为今后顺利开展清水墙修缮及复建施工工作提供了技术保障。

In order to avoid the problems of large interference of human factors,low efficiency of manual marking and low accuracy of visual identification in the traditional manual survey methods,the intelligent identification and evaluation method of typical damage of fair faced wall based on artificial intelligence is studied.Using the methods of deep learning and computer vision,a single-stage target detection neural network model based on YOLO V4 is built,and the training image database of deep learning neural network model for three typical damages of fair faced wall,such as weathering,efflorescence and green plant coverage is established,so as to realize the intelligent diagnosis and regional division of the above three typical damages of fair faced wall.At the same time,based on OpenCV computer vision library,the area ratios of various damaged areas at the pixel size level are calculated to represent the damage degree of fair faced wall.Engineering practice has proved that compared with the traditional manual survey methods,the method based on artificial intelligence has realized an efficient,accurate and convenient automatic identification and rapid assessment process of clear water wall damage,which provides a technical guarantee for the smooth repair and reconstruction of fair faced wall in the future.