本发明公开了一种双视觉飞行机械臂系统,可以有效保证系统同时对静态世界与动态目标的观测质量,提高视觉里程计的鲁棒性与精确性。同时还提出一种动态目标主动追踪与三维重建方法,通过机械臂正运动学建立机体相机与颈眼相机两个独立视觉里程计的联系。同时本发明提出的松耦合式位姿估计系统可以避免机械臂运动学误差对重建精度的影响。对于动态目标的相对位姿估计,提出基于2D PCA与3DDBSCAN的特征分割方法,有效区分动静态特征,提高定位精度。同时提出动态目标的相对静止信息场评估,通过优化最佳探索视角实现动态目标的主动三维重建,有效提高重建效率。基于图像视觉伺服的主动追踪策略,有效减小动态目标丢失的概率。
The invention discloses a double-vision flying mechanical arm system which can effectively ensure the observation quality of the system on a static world and a dynamic target at the same time and improve the robustness and accuracy of a visual odometer. Meanwhile, the invention further provides a dynamic target active tracking and three-dimensional reconstruction method, and the relation between two independent visual odometers of the body camera and the neck-eye camera is established through mechanical arm forward kinematics. Meanwhile, the loose coupling type pose estimation system provided by the invention can avoid the influence of kinematics errors of the mechanical arm on the reconstruction precision. For the relative pose estimation of a dynamic target, a feature segmentation method based on 2D PCA and 3D DBSCAN is provided, the dynamic and static features are effectively distinguished, and the positioning precision is improved. Meanwhile, relative static information field evaluation of the dynamic target is provided, active three-dimensional reconstruction of the dynamic target is achieved by optimizing the optimal exploration view angle, and the reconstruction efficiency is effectively improved. And an active tracking strategy based on image visual servo effectively reduces the probability of loss of the dynamic target.
双视觉飞行机械臂系统及动态目标位姿估计与主动重建方法
2024-07-12
Patent
Electronic Resource
Chinese