To address the issue of non-uniform errors in imaging feature points affecting the accuracy of position calculation in visual navigation,a visual localization method based on variance-covariance component estimation was propose.This method analyzed the existence form of random errors and the influencing factors in the camera measurement process from the theoretical point of view.To create an effective solution,a random model optimi-zation module was integrated into the binocular camera vision and vision/inertial fusion model,resulting in the development of a robust nonlinear optimization system.The algorithm was validated using the KITTI and Eu-RoC datasets,with data processing based on division-based clustering.Experimental results showed substantial improvements in optimizing random errors making it a promising enhancement for visual navigation systems.
关键词
视觉定位/方差分量估计/非线性优化/随机误差/聚类模型
Key words
visual localization/variance component estimation/nonlinear optimization/stochastic model/clus-tering model
维普
万方数据