基于改进SSD模型的道路病害检测研究

Research on Road Damage Detection Based on Improved SSD Model

道路病害的准确检测可以及时定位并修复路面病害。对于保持良好的路况、交通运载、安全驾驶和出行等都具有重要的意义。为了提高现有人工智能道路裂缝检测精度和性能,提出一种基于不规则四边形标注框的改进SSD道路裂缝检测方法。该方法能够在裂缝检测过程中以较高的准确率检测出不同类型的道路裂缝。在建立的不规则四边形标注框数据集上,将改进算法与基于矩形预测框的SSD算法进行比较。实验结果表明,改进的SSD道路裂缝检测方法能够更好的根据裂缝形态进行检测,在检测精度方面有了较高的提升。

Accurate detection of road damage can locate and repair road damage in time.It is of great significance for maintaining good road conditions,traffic carrying,safe driving and travel.In order to improve the accuracy and performance of existing artificial intelligence road crack detection,an improved SSD road crack detection method based on irregular quadrilateral annotation frame is proposed.The method can detect different types of road cracks with high accuracy in the detection process.On the created dataset of irregular quadrilateral boxes,the improved algorithm is compared with the SSD algorithm which is based on rectangular prediction box.The experimental results show that the improved SSD road crack detection method can better detect cracks according to the shape of the cracks,and has a higher detection accuracy.