Observability-Based Single-Camera Placement Optimization for 3D Tracking of Multicopters

Camera placement has a crucial impact on the performance of a three-dimensional (3D) visual tracking system. However, it remains challenging to find the optimal camera placement configuration for 3D tracking of moving multicopters. To address this problem, we propose a single-camera placement optimization method for 3D tracking of moving multicopters from a new perspective — nonlinear observability. First, using the process model for multicopters and a generic camera model, the degree of observability of the extended Kalman filter (EKF) estimator is used as a performance metric and the constraint function is established. Then, the optimal camera placement parameters are obtained by using a genetic optimization algorithm. The proposed camera placement method also has a great application field, which is suitable for conventional cameras as well as fish-eye cameras. Simulation results demonstrate the effectiveness and superiority of the proposed method against the experience-based camera placement configuration for 3D tracking of multicopters.