基于PCNN和遗传算法相结合的新型混凝土桥梁裂缝检测方法

发布时间：2018-03-15 02:18

本文选题：混凝土桥梁裂缝检测　切入点：脉冲耦合神经网络　出处：《计算机应用研究》2017年10期 　论文类型：期刊论文

【摘要】：针对混凝土桥梁裂缝对比度低、裂缝图像噪声干扰强等难题,提出了基于脉冲耦合神经网络(PCNN)和遗传算法相结合的混凝土桥梁裂缝检测新算法(GA-PCNN)。该算法首先利用遗传算法优化裂缝PCNN模型参数;然后通过改进的最小对数误差适应度函数区分裂缝与背景,当适应度值大小几乎无变化时,停止分割图像;最后通过连通域去噪算法滤除残余噪声,实现裂缝的自动检测。比较GA-PCNN、PCNN和基于熵及动态阈值算法对裂缝图像的分割效果,并绘制PR和ROC曲线评价分割质量,经计算GA-PCNN算法的PR和ROC曲线下面积为90.6%和91.6%,分别高于PCNN算法10.1%和6.8%、基于熵和动态阈值6.5%和6.7%。实验结果表明,GA-PCNN新算法分割效果好且去噪能力强,该算法能准确地提取混凝土桥梁裂缝特征。
[Abstract]:Aiming at the problems of low contrast of cracks in concrete bridges and strong noise interference of crack images, A new crack detection algorithm of concrete bridge based on pulse coupled neural network (PCNN) and genetic algorithm (GA) is proposed. Firstly, genetic algorithm is used to optimize the parameters of crack PCNN model. Then the crack and background are distinguished by the improved least logarithmic error fitness function. When the fitness value is almost unchanged, the segmentation of the image is stopped. Finally, the residual noise is filtered by the connected domain denoising algorithm. The effect of GA-PCNNNCN-PCNN and dynamic threshold algorithm on crack image segmentation is compared, and PR and ROC curves are drawn to evaluate the segmentation quality. The area under PR and ROC curves of GA-PCNN algorithm is calculated to be 90.6% and 91.6, which is higher than that of PCNN algorithm 10.1% and 6.8, respectively, based on entropy and dynamic threshold 6.5% and 6.7. the experimental results show that the new algorithm has good segmentation effect and strong denoising ability. The algorithm can accurately extract the crack features of concrete bridges.
【作者单位】：上海理工大学机械工程学院;上海应用技术大学城市建设与安全工程学院;
【分类号】：TP18;TP391.41

【相似文献】