基于改进PCNN模型的机场跑道路面裂缝分割方法  

Segmentation method of airport runway pavement cracks based on improved PCNN model

作　　者：尹欣洁 YIN Xinjie(School of Information Engineering,Hebei GEO University,Shijiazhuang,Hebei 050031,China;Laboratory of AI and Machine Learning,Hebei GEO University,Shijiazhuang,Hebei 050031,China)

机构地区：[1]河北地质大学信息工程学院,河北石家庄050031 [2]河北地质大学人工智能与机器学习研究室,河北石家庄050031

出　　处：《信息记录材料》2022年第6期37-39,共3页Information Recording Materials

摘　　要：针对传统脉冲耦合神经网络(PCNN)对机场跑道路面裂缝检测出现精度低和参数设置复杂的问题,针对PCNN在图像分割时受部分遮挡物、灯光昏暗等影响,出现脉冲误传导、分割效果不理想和分割精度低等问题,提出了一种融合灰度值均方差和最大模糊散度的PCNN图像分割算法。该算法计算单个像素及四周的像素均方差,与相邻的均方差比较,均方差大的作为馈入单元的外部激励输入,有效提高了分割算法的准确度,根据目标图像与背景图像的最大模糊散度,作为迭代终止条件。实验结果表明,光照条件发生变化时对细小裂缝进行完整提取。Aiming at the problems of low accuracy and complex parameter setting of traditional pulse coupled neural network(PCNN) in airport runway pavement crack detection,aiming at the problems of pulse misconduction,unsatisfactory segmentation effect and low segmentation accuracy caused by partial occlusion and dim light in image segmentation,a PCNN image segmentation algorithm integrating mean square deviation of gray value and maximum fuzzy divergence is proposed in this paper.The algorithm calculates the mean square deviation of single pixel and surrounding pixels.Compared with the adjacent mean square deviation,the one with large mean square deviation is used as the external excitation input of the feed unit,which effectively improves the accuracy of the segmentation algorithm.According to the maximum fuzzy divergence between the target image and the background image,it is used as the iteration termination condition.The test results show that the fine cracks can be extracted completely when the light conditions change.

关 键 词：图像分割 脉冲耦合神经网络 模糊散度 机场跑道路面裂缝 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]