【摘要】 针对裂缝图像背景复杂噪声多导致检测精度差的问题，设计了基于特征金字塔注意力的裂缝检测模型(Crack Detection Network Based on Feature Pyramid Attention, FPA)。模型以VGG16为特征提取网络。首先，在模型中引入特征金字塔注意力模块产生像素级注意力，在扩大感受野的同时减少空间分辨率损失的影响。然后，提出多尺度特征融合模块MFF,将具有更大感受野的特征图与每个阶段的特征图进行特征融合，在低层特征中引入语义信息丰富的高层特征，提升对不同尺寸裂缝的检测能力。最后，将全局上下文模块添加到每个分支的特征融合后，提高裂缝区域的权重，减少特征融合过程中低层特征所携带的噪声干扰。通过与多个模型进行对比，所提模型有良好的抗噪性。在CrackTree260数据集上，MPA、MIoU、F-measure指标比其它模型平均提升了3.84、2.09、2.22个百分点。更多还原

【Abstract】 A crack detection model based on feature pyramid attention(FPA-Net) is proposed to address the difficulties in crack detection caused by excessive noise in crack images. The model uses VGG16 for feature extraction. Firstly, feature pyramid attention(FPA) is introduced to expand the perceptual field of the model while generating pixel-level attention to the high-level feature maps. Then, the feature fusion module MFF is proposed to fuse features at each stage with larger perceptual fields feature maps, introducing high-level features into low-level features to enhance the detection of cracks of different sizes. Finally, the global context module(GC) is introduced after feature fusion to enable feature learning to focus on the crack region and suppress the interference information for more comprehensive crack detection. The experimental results show that the model can effectively suppress background noise. On CrackTree260 dataset, the MPA,MIoU and F-measure metrics improved by an average of 3.84、2.09 and 2.22 percentage points over the other models.更多还原