稀疏视觉SLAM对平面激光雷达传感的稠密化模拟

为了解决稀疏特征点VSLAM （visual simultaneous localization and mapping）由于其构图稀疏性，在视觉导航应用方面的短板，提出一种高斯滤波插值的方法对其特征点进行稠密化处理，实现对平面激光雷达反馈的模拟.本文利用高斯分布以及迭代滤波实现数据的稠密化，通过建立全局高斯滤波以及局部高斯分布估计，实现对稀疏VSLAM空间点平面投影的数据插补，进而实现对平面激光雷达数据的模拟.仅使用CPU情况下，算法每帧耗时为0.0003s~0.006s，插值结果相对误差为7.956%.实验证明，该插值方法成功实现了稀疏投影点的稠密化，插补结果与真实激光雷达反馈相似度高，为视觉导航提供了一种有效的前端传感处理方法.

Planar LiDAR Densified Simulation from Sparse Visual SLAM

In order to address the shortcomings of feature-based sparse VSLAM (visual simultaneous localization and mapping) in visual navigation because of its sparse mapping, a densifying algorithm based on Gaussian filter interpolation is proposed to simulate the planar LiDAR feedback. A densifying algorithm based on Gaussian distribution and circulated filtering is proposed. By establishing global Gaussian filter and local Gaussian distribution estimation, the planar projection of the sparse VSLAM spatial points is interpolated to generate a simulated planar LiDAR feedback. With CPU only, the proposed algorithm costs 0.0003s~0.006s for each frame, and gets a 7.956% relative error in its interpolation results. The experiment results demonstrate that the proposed interpolation method can densify the sparse projected points, the result is similar to the true LiDAR feedback, and it provides an effective real-time front-end sensor processing method for visual navigation.