基于自适应邻域Harris算子的三维对象检索方法

针对三维对象检索过程中在对象旋转、灰度改变等复杂情况下检索精度不高的问题,提出一种三维对象检索方法。将Harris算子扩展运用到三维对象,自适应地确定顶点的邻域大小,然后根据Harris函数响应值选取兴趣点。利用兴趣点构建三维对象具有全局形状特征的距离直方图,将距离直方图作为三维形状的描述符进行检索。实验结果验证了算法的有效性,提高了检索的查全率和查准率。

3-D Object Retrieval Method Based on Adaptive Neighborhood Harris Operator

The detection, characterization, and matching of various 3-D features from visual observations is important for a large variety of applications such as modeling, tracking, recognition or indexing. The existing methods detect features by using either photometric information available with geometric information available with 3-D surfaces. To deal with the low retrieval accuracy problem in complex situation of 3-D objects retrieval, such as 3-D objects rotation and brightness changing, and a 3-D objects retrieval method was proposed. The Harris operator was extended the use of 3-D objects, and an adaptive technique was proposed to determine the neighborhood of a vertex. Then the significant interest points were chosen with the Harris response function value. To construct the global shape features distance histogram of 3-D objects with interest points, the distance histogram was used as the 3-D shape descriptor for 3-D object retrieval. Extensive experimental results demonstrated that the proposed method was robust to affine transformations and distortion transformation such as noise addition. Moreover, the distribution of interest points on the surface of an object remains similar in transformed objects, which is a desirable behavior in applications such as shape matching and object registration.