复杂光照场景视觉/IMU稳健定位模型研究

视觉定位技术是综合PNT体系的重要组成部分，复杂光照场景对视觉定位技术鲁棒性带来巨大挑战，严重阻碍视觉定位技术的环境适用性。面向复杂光照场景稳健定位需求，通过研究复杂光照对视觉观测信息的影响机制，建立视觉图像的动态演绎过程、定位影响因子量化表达模型和视觉观测信息光照自适应补偿机制；通过引入低成本IMU定位增强技术，研究复杂光照场景视觉/IMU误差传播与定位精度的耦合关系，构建复杂光照场景视觉观测值误差传播模型，运动载体限制参数和运动信息约束的IMU状态估计模型，视觉/IMU标定参数在线更新模型；顾及复杂光照场景特征，突破光照补偿、抗差自适应定位、视觉定位失效IMU误差补偿技术，构建顾及光照适时补偿的视觉/IMU LSTM-MSCKF稳健定位模型，解决复杂光照场景视觉高精度、稳健定位实现难题。项目成果将丰富视觉定位理论体系，提升视觉定位技术的环境普适性，推动综合PNT体系发展。

Visual positioning technology is an important part of comprehensive PNT system. Complex lighting scenes bring great challenges to the robustness of visual positioning technology, which seriously hinders the environmental applicability of visual positioning technology. In order to meet the needs of robust localization of complex lighting scenes, by studying the influence mechanism of complex lighting on visual observation information, the dynamic deduction process of visual images, the quantitative expression model of positioning influence factors and the adaptive compensation mechanism of visual observation information are established. By introducing low-cost IMU positioning enhancement technology, and studying the coupling relationship between visual/IMU error propagation and positioning accuracy in complex lighting scenes, the error propagation model of visual observations in complex lighting scenes, the IMU state estimation model constrained by motion carrier limiting parameters and motion information, and the online updating model of visual/IMU calibration parameters are constructed. Considering the characteristics of complex lighting scene, breaking through the techniques of illumination compensation, robust adaptive location and IMU error compensation for visual location failure, a vision/IMU LSTM-MSCKF robust positioning model considering timely illumination compensation is constructed to solve the problem of high-precision and robust positioning in complex lighting scene. The results of the project will enrich the theoretical system of visual navigation and positioning, promote the environmental universality of visual positioning technology, and promote the development of comprehensive PNT system.