遥感图像中无水桥梁识别新算法

摘要

给出了边缘密度和线段复杂度的定义,并提出一种遥感图像中无水桥梁的识别新算法.首先对图像进行边缘提取,计算像素点的边缘密度,根据边缘密度进行图像分割,接着采用Hough变换提取直线,利用线段复杂度等确定疑似桥梁区域,然后计算疑似桥梁区域像素点的纹理特征,并构成一个特征矢量,最后将此特征矢量送入BPNN进行分类,统计该区域所有像素点的分类结果以判决是否属于桥梁.实验结果表明,该算法能够较好地识别出遥感图像中的无水桥梁目标.%The definitions of the edge density and the segment complexity are given, and a novel algorithm for recognition of bridge without water under it in remote sensing image is proposed.Firstly,the edge is extracted with Canny operator,and the edge density for each pixel is calculated. The image is segmented base on the edge density. Secondly,the suspected bridge area is identified by a series of processing including using Hough transform to extract straight line and calculating segment complexity. Then texture features of the suspected bridge area are extracted to form a feature vector. Finally, each suspected bridge is classified by Back Propagation Neural Network (BPNN) based on this feature vector. The experimental results show that the algorithm can perform well to detect bridge targets.