基于决策树的路面破损图像快速识别仿真

Rapid Recognition of Pavement Damage Based on Decision Tree

路面破损图像的快速识别方法在提高图像使用率方面具有重要意义。对路面破损图像进行快速识别时,需对破损图像的密度指标进行计算,确定破损图像的结构。传统方法主要根据破损图像的权重进行研究,忽略了图像灰度值的影响,导致破损图像识别率低、实时性差的问题。提出基于决策树的路面破损图像快速识别方法。根据路面破损图像的像素值和图像大小,对路面破损图像的水平和垂直投影进行计算,获得破损图像的特征值,提取图像的边缘值,并计算路面破损图像的边界矩阵、中心边界矩阵和规格化边界矩阵,得到图像的子块总量,计算路面破损裂缝的夹角值,利用破损图像的样本数据和图像的向量转置来确定破损图像的超平面的分类,并对其分类进行优化和求解,判断路面破损图像的类别,计算路面破损图像分类的中心距离,实现对路面破损图像的快速识别。仿真结果表明,提出方法不仅提高了图像的识别率,还具有实时性。

Traditionally, the method mainly ignores the influence of grayscale value of image, resulting in low recognition rate and poor real-time performance. In this paper, we focus on a method to quickly recognize damaged pavement surface image based on decision tree. Based on the pixel value and image size of damaged pavement surface image, we calculated the horizontal projection and vertical projection of damaged pavement surface image and obtained the feature value of damaged image. After that, we extracted the edge value of image and calculated the boundary matrix, the center boundary matrix and the standardizing boundary matrix of damaged pavement surface image, and then we obtained the total number of sub-blocks. Moreover, we calculated the intersection angle value of damaged pavement cracks and used the sample data of damaged image and the vector transposition of image to determine the classification of hyperplane. Meanwhile, we optimized and solved the classification, so as to judge the category of damaged pavement surface image. Finally, we calculated the center distance of classification for damaged pavement surface image. Thus, we achieved the repaid recognition of damaged pavement surface image. Simulation results verify that the proposed method improves the recognition rate of image. Meanwhile, this method has the real-time performance.