针对无特征标志点的大场景多视点云数据，提出了一种新的基于SIFT特征的配准和拼接算法。 算法提出了有效纹理图像的概念，并对有效纹理图像进行SIFT特征提取和匹配；然后将提取的SIFT特征点和匹配关系反射到三维点云数据，获取多视点云数据的特征点和匹配关系，完成多视点云数据的拼接。算法在有效纹理图像中提取和匹配特征点，排除了点云数据中孔洞和无效数据的干扰，并且算法只利用较高鲁棒性的特征点对进行拼接，计算简单，匹配精度和效率都得到提高。对室内和室外两个大场景的2个视点数据进行实验，实验结果证明拼接速度和精度都有较大的提高。

In order to solve multiview point clouds registration in large nonfeature marked scenes, a new registrating and stitching method is proposed based on 2D SIFT(Scaleinvariant feature transform) features. First, a texture mapping method is used to generate 2D effective texture image, from which SIFT features can be extracted and matched; then accurate key points and registration relationship between the effective texture images are obtained. The extracted SIFT key pionts and registration relationship are reflected on the 3D point clouds data to obtain key points and registration relationship of multiview point clouds, so the multiview point clouds stitching can be completed. In this method, the interference of the hole and ineffective points can be eliminated by using the texture mapping method. Experiments on twoview point clouds in indoor and outdoor scenes are carried out and the results prove that the matching precision and efficiency are greatly improved.