基于不变矩特征的沥青路面破损图像识别 |
| |
引用本文: | 初秀民,王荣本,储江伟,王超. 基于不变矩特征的沥青路面破损图像识别[J]. 吉林大学学报(工学版), 2003, 33(1): 1-7 |
| |
作者姓名: | 初秀民 王荣本 储江伟 王超 |
| |
作者单位: | 吉林大学,交通学院,吉林,长春,130025 |
| |
基金项目: | 国家教育部博士学科点基金资助项目(2000018507) |
| |
摘 要: | 提出了一种减少沥青路面破损图像识别计算量的图像分割方法。将路面图像等分为64×64像素的子块图像,并用灰度方差值描述子块图像特征。设计了基于BP神经网络的子块图像模式分类器,利用子块图像模式分类结果所组成的矩阵作为路面破损图像分割结果。并将路面破损图像子块模式矩阵的不变矩作为路面破损图像的整体特征,在此基础上设计了基于全局优化算法的路面破损前馈神经网络分类器。最后进行了路面破损图像识别试验,识别率达到83 3%。
|
关 键 词: | 路面破损 模式识别 不变矩 神经网络 |
文章编号: | 1671-5497(2003)01-0001-07 |
修稿时间: | 2002-05-13 |
Asphalt Pavement Surface Distress Image Recognition Based on Moment Invariant Feature |
| |
Abstract: | A method of image feature representation is put forward,which may reduce calculation of pavement surface distress image classification.Pavement surface image is divided into 64×64 pixels subimages,and greyness variances are used to represent subimages feature.Meanwhile,a subimage pattern classifier is designed based on BP artificial neural network,all classifying results of subimages pattern are arrayed in a matrix,which represents pavement surface distress image segmentation.Moment invariant of matrix is used to represent pavement distress image feature,and furthermore,a pavement distress forward feed neural network classifier is designed based on global optimization algorithm.Finally,pavement distress image recognition experiment is done,and accurately classifying rate is 83.3%. |
| |
Keywords: | pavement surface distress model recognition moment invariant neural network |
本文献已被 CNKI 万方数据 等数据库收录! |
|