融合注意力机制和区域生长的裂缝识别算法研究

Crack Recognition Algorithm Integrating Attention Mechanism and Region Growth Method

为了优化传统裂缝识别方法,提升裂缝识别方法的鲁棒性和识别精度,实现自动化裂缝识别。提出了一种融合注意力机制和区域生长的裂缝识别算法,建立动态阈值,通过区域生长方法获取动态阈值下的裂缝结果,引入注意力机制,根据动态阈值下裂缝识别结果为图像各像素点赋予不同的权重系数,得到较准确的裂缝识别结果,最后通过裂缝骨架长度和裂缝区域所占面积确定裂缝的特征信息,即裂缝长度和裂缝平均宽度。通过现场不同裂缝图像进行验证,研究结果表明:融合注意力机制和区域生长的裂缝识别方法识别出的裂缝特征信息与裂缝实际特征基本吻合,具有较强的鲁棒性和识别精度,可为工程健康监测提供支持。

To optimize traditional crack identification methods,improve the robustness and accuracy of the crack identification method,and realize automatic crack identification,this paper proposes a crack recognition algorithm integrating attention mechanism and region growth.It establishes dynamic thresholds and obtains the crack results under different thresholds by employing the region growth method.The attention mechanism is introduced to assign different weight coefficients to each pixel of the image according to the crack identification under dynamic thresholds for more accurate crack recognition results.Finally,the characteristic information of cracks is determined by the length of the crack skeleton and the area occupied by the crack region(length and average width of the crack).Through verifying field crack images,the results show that the crack characteristic information identified by the crack identification method integrating attention mechanism and region growth is basically consistent with the actual crack characteristics.The method features strong robustness and identification accuracy,which provides support for engineering health monitoring.