针对道路裂纹图像识别准确率低的问题，提出了一种基于视觉注意力机制的道路裂纹识别方法。首先人工将源图像分类为无裂纹、纵向裂纹、横向裂纹、网格裂纹。再通过对裂纹图像进行Gabor滤波获得相应纹理信息，对获得的纹理图像进行K-means聚类算法实现图像分割。然后结合基于注意力机制的模型SRMResnet（Style-based Recalibration Module Of Resnet）进行分类识别。最后将提出的方法与几种常见基于注意力机制的模型在4类图像上进行对比测试。结果显示，基于改进视觉注意力机制的道路裂纹识别对分类能力的准确率有较大提高，能够高效识别道路裂纹，最终的识别准确率可以达到0.966。

Aiming at the problem of low accuracy of road crack identification, a road crack identification method based on visual attention mechanism is proposed. Firstly, the source images are manually classified into non-crack, longitudinal crack, transverse crack, and grid crack. Then the corresponding texture information is obtained by Gabor filtering of the crack image, and the obtained texture image is segmented by K-means clustering algorithm. Then the attention-mechanism-based model SRM-Resnet (Style-based Recalibration Module Of Resnet) is used for classification and recognition. Finally, the proposed method is compared with several common attention-based models on four types of images. The results show that the road crack identification based on the improved visual attention mechanism proposed in this paper has greatly improved the accuracy of classification ability, and can effectively identify road cracks. The final identification accuracy can reach 0.966.