Abstract: It is foundation for automatic inspection to collect pavement image and do the image binarization.In this paper，a high resolution image collecting system is built to real-time collect and save pavement images，and the pavement marker image is departed from the background of pavement image by mean of analysing the pixels grey.In pavement image binarization，the pavement image is divided into 400 sub-images which have 100×100 pixels.First，the features of sub-images are represented by the aij and bij，and the sub-image model classifier is designed by mean of online study.Second，the connexity of the sub-image model is used to describe pavement surface distress continuity，farther the wrong classifying models of the pavement sub-images are corrected.Last，the optimization threshold of the pixels grey is used to do the pavement surface image binarization.The test shows that this arithmetic can segment the pavement surface distress clearly.

Key words: pavement management system, pavement surface distress, image recognition, binarization

摘要： 路面图像采集和二值化是路面破损自动分析的基础。构建了高分辨路面图像自动采集系统，实时采集、存储路面图像，采用直方图分析法将道路标线从路面图像的背景中去除。在路面图像二值化中，采用分块处理策略。首先利用子块图像灰度方差与标准子块好路面图像灰度方差以及该子块图像灰度方差均值的比值描述路面子块图像特征，并采用在线学习的方法设计路面子块图像模式分类器；其次利用子块图像模式的连通性，有效描述路面破损在空间上的连续性，进而去除路面破损图像误判的子块模式；最后采用最佳阈值法将所有路面破损子块图像二值化，实现整幅路面图像的二值化。实验结果表明，该文算法可有效分割路面破损图像。

关键词: 路面管理系统, 路面破损, 图像识别, 二值化