Visual place recognition is an important issue in both computer vision and robotics.Changes in image content caused by viewpoint changes in natural scenes still pose a challenge to location recognition.To solve this problem,a novel feature reorganization method based on location clustering is proposed.Firstly,a general pyramid expansion scheme is extracted based on histogram features,called PyramidV-LAD.To maximize the effect of the new function,the similarity is evaluated by clustering images with a certain threshold into same location.Extensive experiments have been conducted to verify the effective-ness of the proposed method using Pyramid VLAD to compare with the best method,APANet.These two datasets achieve improvements of 1.02 and 2.54 percent points in Recall@1,respectively.The results show that this method can consistently obtain better performance than the state-of-the-art methods on the two standard place recognition benchmarks.
visual place recognitionpyramid principal phasesplace clusteringimage processing
国家科技期刊平台
NSTL
万方数据
