A review of deep learning methods for tunnel lining leakage identification
Lining leakage is one of the major diseases in tunnel engineering,and there are various problems such as long time consumption,low efficiency,poor reliability,strong influence from external conditions,and serious distance attenuation in the identification of tunnel lining leakage sites by traditional methods. Deep learning has made great progress in the field of computer vision,which can well accomplish a series of image processing tasks and can reliably perform target detection of tunnel lining diseases by identifying tunnel lining images. This study reviewed the research on the application of deep learning models in the task of tunnel lining leakage identification from three aspects of image classification,target detection,and image segmentation,analyzed the advantages,disadvantages,and inheritance relationship of each model,and evaluated the performance of each model by combining the indexes of precision,recall,F1,mAP and mIoU. It finally summarized the shortcomings of the existing deep learning models and gave a targeted outlook on the future development of the models,so as to provide a basis for improving the effectiveness of tunnel lining leakage identification and promote the further application of deep learning in this field.
deep learningimage classificationtarget detectionimage segmentationtunnel leakage
万方数据
维普
