基于图象子块分布特性的路面破损图象特征提取

由于路面破损形式的多种多样,造成路面破损分类1]成为一大难题,这极大的限制了路面破损自动检测的普及和发展,使得路面破损自动检测即使在发达国家也普及得不够理想。本文在前文提出的破损密度因子的基础上,进一步设计了出方向密度因子,得到一种基于图象子块分布特性的路面破损识别算法。通过仿真,验证了其对常见的5种路面破损类型进行分类的可行性。为了进一步验证我们提出的识别算法,论文选择了另外一种路面破损分类算法,即PROXIMITY算法进行神经网络仿真对比。神经网络的训练样本是两组,测试样本也是两组,进行了四次仿真对比。四次仿真结果都显示方向密度因子算法优于PROXIMITY算法。

A STUDY ON PAVEMENT SURFACE DISTRESS IMAGE FEATURE EXTRACTION BASED ON DISTRIBUTIVE CHARACTERISTICS OF IMAGE TILES

Conventional visual and manual road crack detection method is no fit to the development of highways: it is labor-intensive,time-consuming,imprecise,dangerous and costly,and also is can affect transportation.Therefore,automatic pavement survey is required.Mainly because there are too many types of cracks,most advanced equipment still has difficulty to classify cracks,even though they are easy to detect.Automatic pavement crack classification based on image processing is studied,and a new method based on DIRECTIONAL DENSITY FACTOR for pavement distress classification is put forward in this paper.In order to improve the accuracy and efficiency to identify the pavement surface distress by the image information,a method of pavement surface image feature representation is used,which can reduce the amount of calculation of pavement surface distress image classification.A pavement surface image is divided into 64x64 pixels sub-images,all of the sub-image pattern classifying results are arranged into a matrix,and the pavement surface distress image feature is represented by this matrix.Simulation indicates that the DIRECTIONAL DENSITY FACTOR algorithm can classify five most common types of cracks(longitudinal crack,transverse crack,block crack,alligator crack and no crack) very well.At the same time,four simulations show that the DIRECTIONAL DENSITY FACTOR algorithms are better than the PROXIMITY algorithm.