当前位置: X-MOL 学术Image Vis. Comput. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Intelligent detection of building cracks based on deep learning
Image and Vision Computing ( IF 4.7 ) Pub Date : 2020-07-30 , DOI: 10.1016/j.imavis.2020.103987
Minjuan Zheng , Zhijun Lei , Kun Zhang

In order to solve the damage caused by the concrete structure, which leads to the reduction of the life of infrastructure, endangers the safety of pedestrians, and has a serious impact on the social economy, building crack detection model of FCN (Fully Convolutional Networks), R-CNN (Regions with CNN feature) and RFCN (Richer Fully Convolutional Networks) has been proposed based on the convolutional neural network model to amplify and extract the features of the data and previous studies. Through the training of building surface data such as roads, bridges, houses and dams, the model is analyzed in terms of morphological and geometric indexes. Finally, the model of crack picture detection and segmentation based on deep learning is used for picture performance detection and comprehensive evaluation. The results show that: in the aspect of building gap detection, the RFCN model has the best processing effect, the gap recognition degree is higher, and the detail processing is better. In the aspect of model evaluation index, the correct rate of RFCN model is increased by 10%, the accuracy rate is increased by 12%, the recall rate is increased by 8%, the loss rate is increased by 3%, and the overall stability is higher. In the aspect of comprehensive performance, the picture processing performance is better than the FCN model by 7% and better than the R-CNN model by 15%, and the memory share is 80%. The fusion model based on deep learning and picture processing has been improved in many aspects, which can provide strong theoretical support and practical value for the detection and research of concrete surface cracks such as bridges, dams, highways and houses.



中文翻译:

基于深度学习的建筑裂缝智能检测

为了解决混凝土结构造成的破坏,从而导致基础设施寿命的减少,危害行人的安全,并对社会经济产生严重影响,建立了FCN(Fully Convolutional Networks)裂缝检测模型基于卷积神经网络模型,已经提出了R-CNN(具有CNN特征的区域)和RFCN(Richer完全卷积网络)以放大和提取数据的特征以及先前的研究。通过训练诸如道路,桥梁,房屋和大坝等建筑表面数据,从形态和几何指标方面对模型进行了分析。最后,将基于深度学习的裂纹图像检测与分割模型用于图像性能检测和综合评价。结果表明:在建筑物缝隙检测方面,RFCN模型具有最佳的处理效果,缝隙识别度较高,细节处理效果更好。在模型评估指标方面,RFCN模型的正确率提高了10%,准确率提高了12%,召回率提高了8%,丢失率提高了3%,总体上稳定性更高。在综合性能方面,图像处理性能比FCN模型好7%,比R-CNN模型好15%,内存份额为80%。基于深度学习和图像处理的融合模型已经在很多方面进行了改进,可以为桥梁,大坝,公路和房屋等混凝土表面裂缝的检测和研究提供强大的理论支持和实用价值。

更新日期:2020-07-30
down
wechat
bug