把 SIFT 算法应用在牙齿模型图像上，检测牙齿图像的特征点。 方法：首先采用高斯差分算子 DoG 搜索整个图

像的尺度和位置信息，从而确定具有代表性尺度、方向的特征点。基于其稳定性选择关键点，得到一个详细的模型以确定每个候

选点的合适位置和范围。基于局部图像梯度方向信息将方向矢量和关键点对应起来。在选定范围内的每个关键点周边区域测量

局部图像梯度，并采用 KNN 算法进行特征匹配。 结果：通过大量的实验和与其他特征提取方法相比较，该方法能有效地检测牙

齿模型图像的特征，并为牙齿模型三维重建提供有效的参数。-SIFT algorithm is applied to the teeth of the model image, the image feature point detecting teeth. Methods: DoG Gaussian differential operator to search the entire image the scale and location information, to determine a representative scale, the direction of the feature point. Select the key points based on their stability, to get a detailed model to determine the appropriateness of each candidate point location and extent. Information based on local image gradient direction and key points of the direction vectors correspond. Within the selected area around each critical point of measuring the local image gradient, and using KNN algorithm for feature matching. Results: Through a lot of experiments with other feature extraction methods and compare the proposed method can effectively detect tooth model image feature, and to provide an effective three-dimensional reconstruction tooth model parameters.