Object-Aware SLAM Based on Efficient Quadric Initialization and Joint Data Association

Semantic simultaneous localization and mapping (SLAM) is a popular technology enabling indoor mobile robots to sufficiently perceive interact with the environment. In this paper, we propose an object-aware semantic SLAM system, which consists of quadric initialization method, object-level data association multi-constraint optimization factor graph. To overcome limitation multi-view observations requirement dense point clouds for objects, efficient method based on object detection surfel construction proposed, can efficiently initialize quadrics within fewer frames small viewing angles. The robust joint tightly coupled graph bundle adjustment enable accurate estimation constructed camera poses. Extensive experiments using public datasets show that proposed system achieves competitive performance respect accuracy robustness compared state-of-the-art methods.