Fast k-Nearest Neighbor Searching in Static Objects

The k-nearest neighbor searching is a classical problem that has been seriously studied, due to its many important applications. The paper proposes an efficient algorithm to search the k-nearest neighbors for static objects. Since locations of static objects are known in advance and not changed, most of existing solutions build a kd-tree as a preprocessing and search the k- nearest neighbors by using it. We propose a completely different preprocessing with kd-trees. The core idea of this paper is to build in advance the k-nearest neighbors of each static object as a preprocessing. If a querying point q is given, the nearest object p of q is firstly searched and then the k-nearest neighbors of q are found by using the k-nearest neighbors of p. It is to use the feature that two objects may share many neighbors if they are spatially close to each other. In order to measure the performance of the proposed algorithm, we have a number of experiments. The results of experiments showed that the proposed algorithm is 2-3 times quicker than the method using kd-tree in the Point Cloud Library(PCL).