本文研究的主要目的是如何实现高分辨率遥感影像道路目标的自动识别与提取。在研究过程中,首先采用各种遥感影像分割算法来分割出道路信息，而后使用数学形态学以及模式识别相关知识来实现道路目标的自动提取。本文的主要内容有： （1）在对影像进行分割之前首先进行预处理，预处理对分割效果将会产生较大影响，尤其在影像增强方面，所以本文总结了一些较为成功的影像增强方法，并分别采用遥感影像直方图均衡化以及数学形态学方法对影像进行增强处理； （2）本文对高分辨率遥感影像Canny边缘检测算法、阈值分割算法、K均值聚类分割算法等进行了重点研究，并对其进行了改进，以适合高分辨率遥感影像道路特征自动提取的条件； （3）实践证明，任何一种分割方法都很难得到令人满意的分割效果，而将某些分割方法融合应用能取得较好的分割结果。因此，本文将某些分割方法联合应用，研究了将边缘检测分割算法与K均值聚类分割算法相结合来提取道路信息的方法，实验证明能够取得较好的分割结果； （4）得到较好的分割结果后，本文应用数学形态学影像处理方法对分割结果进行了进一步的处理，而后应用模式识别相关知识识别道路目标，得到了较好的道路信息。

The main purpose of this article is how to achieve the automatic recognition and extraction of road target from high-resolution remote sensing images. In the process of research, first uses all sorts of algorithms of remote sensing images segmentation to segment the road, then uses the relevant knowledge of mathematical morphology and pattern recognition to achieve automatic extraction of the road target. The main content of this article is: (1) First do pre-processing before the segment of images, pre-processing will have great influence on segmentation effect (particularly in image enhancement). This article summarizes some successful methods of the image enhancement, and for the image’s enhancement processing by using remote sensing image histogram equalization and method of mathematical morphological. (2) This paper focuses on the Canny edge detection algorithm、threshold segmentation algorithm、K-means clustering segmentation algorithm etc, and improved them, so for meeting the conditions of automatic extraction of road features from high-resolution images. (3) Practice has proved, any kind of segmentation methods are difficult to get satisfactory segmentation results, however when some segmentation methods fusion applied can obtain good results. Therefore, this paper combined some segmentation methods to application, the research studies the method of combining edge detection segmentation algorithm and K-means clustering segmentation algorithm to extract the road information. (4) The good segmentation results have obtained, using the method of mathematical morphology image processing for further processing of the segmentation results, then using the knowledge of pattern recognition to recognize road target, getting a good road information.