2018/03/02 by Pilailuck Panphattarasap, Panphattarasap, Pilailuck, Andrew Calway +1 · 2 citations
Computer Science · Engineering · #Advanced Image and Video Retrieval Techniques #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Indoor and Outdoor Localization Technologies #Robotics (cs.RO) #Robotics and Sensor-Based Localization
paper · pdf · doi:10.48550/arxiv.1803.00788
openalex publication_date 2018/03/02 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
We describe a novel approach to image based localisation in urban\nenvironments using semantic matching between images and a 2-D map. It contrasts\nwith the vast majority of existing approaches which use image to image database\nmatching. We use highly compact binary descriptors to represent semantic\nfeatures at locations, significantly increasing scalability compared with\nexisting methods and having the potential for greater invariance to variable\nimaging conditions. The approach is also more akin to human map reading, making\nit more suited to human-system interaction. The binary descriptors indicate the\npresence or not of semantic features relating to buildings and road junctions\nin discrete viewing directions. We use CNN classifiers to detect the features\nin images and match descriptor estimates with a database of location tagged\ndescriptors derived from the 2-D map. In isolation, the descriptors are not\nsufficiently discriminative, but when concatenated sequentially along a route,\ntheir combination becomes highly distinctive and allows localisation even when\nusing non-perfect classifiers. Performance is further improved by taking into\naccount left or right turns over a route. Experimental results obtained using\nGoogle StreetView and OpenStreetMap data show that the approach has\nconsiderable potential, achieving localisation accuracy of around 85% using\nroutes corresponding to approximately 200 meters.\n