2018/04/10 by Lachlan Nicholson, Michael Milford, Nicholson, Lachlan +3 · 6 citations
Computer Science · Engineering · #Advanced Image and Video Retrieval Techniques #Advanced Neural Network Applications #FOS: Computer and information sciences #Robotics (cs.RO) #Robotics and Sensor-Based Localization
paper · pdf · doi:10.48550/arxiv.1804.04011
openalex publication_date 2018/04/10 · openalex created_date 2022/10/03 · openalex updated_date 2026/07/28
In this paper, we use 2D object detections from multiple views to\nsimultaneously estimate a 3D quadric surface for each object and localize the\ncamera position. We derive a SLAM formulation that uses dual quadrics as 3D\nlandmark representations, exploiting their ability to compactly represent the\nsize, position and orientation of an object, and show how 2D object detections\ncan directly constrain the quadric parameters via a novel geometric error\nformulation. We develop a sensor model for object detectors that addresses the\nchallenge of partially visible objects, and demonstrate how to jointly estimate\nthe camera pose and constrained dual quadric parameters in factor graph based\nSLAM with a general perspective camera.\n