在计算机辅助近景摄影测量系统中，用相机从不同方位对场景拍摄一组照片，在计算机上对该组照片上的目标点进行标记，获得各个标记点在不同照片中的二维坐标值，经过一系列算法，将二维坐标转化为三维坐标，从而完成整个测量过程，而照片中点与点的匹配是这一系列算法中必不可少的一步。 所谓匹配，就是在两幅或者多幅图像中找出对应同一对象的标记点。本文图像点匹配方法分三步实现。首先，标记点的识别和定位。目前，识别方法多种多样，本文详细描述了一种改进后的识别方法，并对其几个步骤做了详细阐述，包含二值化图像、边缘提取和扫描判断。实验结果证明，该方法可以高效的识别标记点。 其次，是对识别出来的编码点进行解码和匹配。本文详细的描述了编码点的编码设计步骤，以及依据编码原理的唯一性，实现解码和匹配。有了第一步的准备，第二步变的计算简单、容易显现。实验结果表明，该方法可行，匹配率极高，值得借鉴。 最后，本文对非编码点的匹配做了大量详细深入的研究，通过对比几种非编码点的匹配方法，提出一种改进后的匹配方法。由于非编码点没有编码点的特定外形和其庞大的数量，本文将采用编码点和非编码点相结合的方法，并引入相似性和摄影几何学中的不变定理，对非编码点完成了匹配。实验结果表明，该方法匹配效果良好，有一定的使用价值。

In the computer-aided close-range photogrammetry, a group of photos from differert directions with the camera, in the computer mark the target point from the group of photos, obtain two-dimensional coordinates of each marker, after a series of algorithms, two-dimensional coordinates into three-dimensional, and complete the whole measurement process, and the point and point matching is an essential setp in this series of algorithms. The matching, that find the makers corresponding to the same objiect in two or more images. This image point matching method will be realized in three steps. Firstly, the marked point is to be locationed and identified. Currently, there is more and more identification methods, this article describes an improve identification method in detail. Each step has been discussed in detail, including image binarization, edge detection and scanning. Experimental results show that the method can efficiently identify the identifying point. Secondly, the coded point is identified and it is to be decoded and matched. This article describes the design steps of the encoding code point in detail, and uniqueness based on coding theory, achieve to decode and match. With the first step to prepare, this step change calculation is simple, easy to show. Experimental results show that the method is feasible, matching high, it is worth drawing. Finally, there are a lot of study on the match of non-coded point. By comparing several non-coded point matching method, proposed an improve method for the match. Due to non-coded point lack of the characteristics of coded point shape and large numbers, we use method of combining coding and non-coding point, and the introduction of similarity and Geometry theorems in the same photo, completed the match of non-coding point. The results show that the method has a good match results and certain value.