Rich Common Crucial Feature for Crowdsourcing-Based Mobile Visual Location Recognition

Crowdsourcing provides an effective way to construct a location recognition image database. Comparing with traditional location image database construction, crowdsourced image database has massive advantages, e.g., much richer information for location, with various angles, timestamps, distances and weather information, providing useful potential for high recognition precision. However, when capturing the crowdsourced images, it is inevitable to have various disturbances on these location images, for example, moving vehicles and pedestrians, hindering the realization of potential. To address this challenge, we first propose a Rich Common Crucial Feature (RCCF) detection framework to exclude unimportant visual SURF features from crucial features. To achieve a good balance between the efficiency and accuracy, we further propose an RCCF based Visual Hash Bits (VHB) scheme to encode RCCF features into hash bits to vote for most matching images. Furthermore, deep feature extraction is also utilized with visual search architecture MobileNet. Extensive experiments are conducted on a crowdsourced dataset with 9,064 location images, demonstrating that our scheme outperforms other state-of-the-art schemes.