Fast loop closure detection via binary content

• Main Authors: Wang, Han, Li, Juncheng, Ran, Maopeng, Xie, Lihua
• Other Authors: School of Electrical and Electronic Engineering
• Format: Conference or Workshop Item
• Language: English
• Published: 2020
• Subjects: Engineering::Electrical and electronic engineering; Loop Closure; Binary Content
• Online Access: https://hdl.handle.net/10356/138248
• Institution: Nanyang Technological University

Loop closure detection plays an important rolein reducing localization drift in Simultaneous LocalizationAnd Mapping (SLAM). It aims to find repetitive scenes fromhistorical data to reset localization. To tackle the loop closureproblem, existing methods often leverage on the matching ofvisual features, which achieve good accuracy but require highcomputational resources. However, feature point based methodsignore the patterns of image, i.e., the shape of the objects aswell as the distribution of objects in an image. It is believedthat this information is usually unique for a scene and can beutilized to improve the performance of traditional loop closuredetection methods. In this paper we leverage and compressthe information into a binary image to accelerate an existingfast loop closure detection method via binary content. Theproposed method can greatly reduce the computational costwithout sacrificing recall rate. It consists of three parts: binarycontent construction, fast image retrieval and precise loopclosure detection. No offline training is required. Our methodis compared with the state-of-the-art loop closure detectionmethods and the results show that it outperforms the traditionalmethods at both recall rate and speed.