Object Detection Based on Multiresolution CoHOG

摘要: Recently, co-occurrence histograms of oriented gradients (CoHOG), a method for describing image features in order to calculate the pixels allocated at local level, has attracted attention as an effective object detection. However, there are some problems. For feature descriptions that focus on individual pixels, calculation cost and number dimensions tends increase exponentially with pixels. This paper proposes multiresolution (MRCoHOG) description is able suppress these exponential increases into linear without reducing classification accuracy. MRCoHOG can reduce by calculating only between adjacent it maintains accuracy extracting from multiple low-resolution images. We performed experiments using vehicle data set cropped surveillance images parking area INRIA Person Data Set, results showed performance equivalent CoHOG.