刑事犯罪现场的图像拼接设计与实现

在刑事犯罪现场辅助测量中，图像拼接对于大面积现场来说显得至关重要。本文提出了一种新的图像拼接算法，它根据两幅现场图像样本之间的一种简单的对应点关系，计算图像间的变换矩阵，通过变换矩阵建立从不同拍摄平面到统一的世界坐标的反投影，在统一的世界坐标下使样本图像之间的匹配误差达到最小。这种变换关系保证了图像变换
换平移不变性和规模不变性，从而保证了全局优化的结果。在图像整合中，本算法改进了传统图像整合中“相似距离”的定义，提出了一种“极值点”算法，大大地降低了计算复杂度，使算法具有较高的性能。该方法已用于刑事犯罪现场图像的拼接，并取得了较为满意的效果。

Design and Implementation of Image Mosaics in the Criminal Scene

In the computer-aided measurement in criminal scenes, image mosaics is very important for large-scale scenes. In the paper, the give a new method of image mosaics, which includes computing the homograph of them according to the simple corresponding points between the two scene images and establishing anti-projection transform from all sorts of photograph planes to the uniform world coordinate by a transform matrix. The algorithm reduces the sample image matching error. The transform assures the invariability of the image's moving and scale, and obtains the optimal solution. In the image registration process, It brings forward a new method named “extremum point”. The method greatly reduces the computing complexity and improves the capability. The algorithm has been applied to criminal scenes with satisfactory results.