基于遗传神经网络的机场跑道局部影响的裂缝检测算法  

Crack Detection Algorithm Based on Genetic Neural Network for Runway Local Influence

作　　者：陈俣秀[1] 高建民 贺青青 高建国 CHEN Yuxiu;GAO Jianmin;HE Qingqing;GAO Jianguo(School of Transportation Science and Engineering,Civil Aviation University of China,Tianjin 300300,China;School of Science,Chang an University,Xi’an 710064,China)

机构地区：[1]中国民航大学交通科学与工程学院,天津300300 [2]长安大学理学院,西安710064

出　　处：《计算机测量与控制》2023年第5期7-13,共7页Computer Measurement &Control

基　　金：天津市科技发展计划项目(2019-18)。

摘　　要：针对机场跑道裂缝的自主识别和提取过程中存在的阴影、光照不均匀以及效率和精度难以兼顾等一系列问题,提出利用遗传算法优化神经网络的机场道面裂缝检测算法;首先,将拍摄的机场道面裂缝图像进行预处理,包括图像灰度化、高斯滤波以及ROI区域确定;设定神经网络拓扑结构,初始化编码长度以权值阈值及等参数,利用选择、交叉和变异等操作反复执行至最佳进化解,进而搭建匹配的神经网络,获得最大分割阈值;结果表明,遗传神经网络算法在综合评价、召回率和准确率3个评价指标上均具有显著提升,其均值分别为93.22%、96.28%、90.75%,实现了在复杂背景下对裂缝提取的目标,为机场道面的后期维护和保养提供了技术支持。Aimed at the problems of shadow,non-uniform illumination,difficultly balancing efficiency and accuracy in the process of automatic crack identification and extraction of airport runway,on the basis of neural network of genetic algorithm optimization,an airport runway crack detection algorithm is proposed.Firstly,the algorithm preprocesses the airport pavement crack images,including image graying,Gaussian filtering and determination in ROI region.Secondly,by setting the network parameters of the genetic algorithm,the parameters of encoding length and weight threshold are initialized,the operations of selection,crossover and mutation are repeatedly executed to the optimal progressive solution.Finally,a matching neural network is built to obtain the maximum segmentation threshold.The results show that the genetic neural network algorithm has a significant improvement in the comprehensive evaluation,recall rate,and accuracy of three evaluation indexes,their average values are 93.22%,96.28%,90.75%,respectively,which achieves the target of crack extraction under a complex background,it provides technical support for the later maintenance and maintenance of airport pavement.

关 键 词：机场道面 遗传神经网络 裂缝检测 特征提取 局部阴影 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TM73[自动化与计算机技术—控制科学与工程]