基于计算机视觉的公路边坡裂缝检测方法

裂缝是大部分公路边坡灾害的早期症状,安全监测集中在此阶段进行.目前,公路边坡多采用人工定期巡查等方法,但存在成本高、监测范围小、人工干预多、安全性差等问题.提出一种基于计算机视觉技术的边坡裂缝监测技术,使用专业级摄像头拍摄边坡裂缝图片结合人工标注,构建目前种类多样的、且符合标准的边坡裂缝数据集;基于此数据集,利用深度学习、膨胀卷积等思想设计了边坡裂缝检测模型FSNet,实现了裂缝的精准分割与识别.经实验证明,该模型对边坡裂缝具有较好的识别能力,识别准确率达到94.21％,且该模型网络参数少、运算复杂度低,为实现公路边坡智能化监测提供可行性.

Highway Slope Crack Detection Method Based on Computer Vision

Crack is the early symptom of most highway slope hazards,and safety monitoring is concentrated in this stage.At present,regular manual inspection and other methods are often used for highway slope,but there are some problems,such as high cost,small monitoring range,more manual intervention and poor safety.A kind of slope crack monitoring technology based on computer vision technology is proposed.By using professional-grade camera to take slope crack pictures and manual annotation,various and standard slope crack data sets are constructed at present.Based on this data set,the slope crack detection model FSNet with the ideas of deep learning and expansion convolution is designed,so as to achieve accurate fracture segmentation and identification.The experimental results show that the model has a good ability to identify slope cracks,and the identification accuracy reaches 94.21%.Moreover,the model has few network parameters and low computational complexity,which provides feasibility for realizing intelligent monitoring of highway slope.