基于深度学习方法的大坝裂缝检测研究

Research on Dam Crack Detection Based on Deep Learning Method

由于混凝土材料力学性质和地质环境的影响,水库蓄水后大坝有可能出现裂缝导致坝渗漏,影响水库发电、供水、灌溉等效益的正常发挥。因此,提前对混凝土进行裂缝识别具有较高的应用价值和科学意义。在卷积神经网络基础上采用语义分割DeepCrack裂纹检测模型,以某水电站附近区域裂缝为输入数据集,结合超高分辨率图像进行裂缝自动检测研究。研究结果表明,基于深度学习的图像处理方法可以自动提取混凝土表面的损伤区域,其精度高于人工检测,但图像处理方法的精度受输入图像质量的影响。此外,拍摄条件是决定裂缝识别精度的因素之一,评价拍摄条件的有效性对于提高裂缝识别方法的可靠性非常重要。

Due to the influence of Strength of materials properties of concrete and geological environment,cracks may appear in the dam after the impoundment of the reservoir,which may lead to dam leakage,affecting the normal play of power generation,water supply,irrigation and other benefits of the reservoir.Therefore,early identification of cracks in concrete has high application value and scientific significance.On the basis of Convolutional neural network,the semantic segmentation DeepCrack crack detection model is adopted,and the cracks in the vicinity of a hydropower station are taken as the input data set,combined with ultra-high resolution images to carry out automatic crack detection research.The research results indicate that image processing methods based on deep learning can automatically extract damage areas on concrete surfaces.Its accuracy is higher than manual detection,but the accuracy of image processing methods is affected by the quality of the input image.In addition,shooting conditions are one of the factors that determine the accuracy of crack identification,and evaluating the effectiveness of shooting conditions is crucial for improving the reliability of crack identification methods.