基于历史信息的路面表观损坏图像识别

Pavement Distress Detection Based on Historical Information

摘要:

在路面表观损坏图像识别方法中，为了解决现有方法准确度与速度不高的问题，利用历史检测结果这一历史信息，提出一种基于历史信息的路面表观损坏图像识别方法。首先，搭建了面向历史信息应用的算法框架，引入利用历史信息创建损坏识别的初始约束条件的机制。其次，训练VGG-16网络用于提取损坏特征。最后，利用历史信息建立初始种群以及设计特征参数，从而改进遗传算法。实验结果表明，该方法相较不运用历史信息的识别方法，能在不降低识别准确度的前提下显著提升识别速度，最快可较不运用历史信息的算法提升141.71倍速度。

Abstract:

In order to solve the problem that the accuracy and speed of the existing methods are not high in recognition of pavement distress image， a recognition method of pavement distress image based on historical information is proposed by using historical detection results. First， an algorithm framework for the application of historical information is built， and the mechanism of using historical information to create initial constraints for distress detection is introduced. Next， the VGG-16 network is trained to extract damage features. Finally， the initial population is established， and the feature parameters are designed by using historical information to improve the genetic algorithm. The experimental results show that this method can significantly improve the recognition speed without reducing the recognition accuracy compared with the recognition method without using historical information， and the fastest speed can be 141.71 times higher than the algorithm without historical information.