无人车基于激光雷达/视觉的目标体积自动测量方法

近些年，基于激光雷达和视觉的目标感知在无人系统中得到了广泛应用。目标的体积测量在很多应用场景可以发挥极其重要的作用，然而对识别感知目标的体积测量，目前尚无大量研究。首次提出了一种基于激光雷达/视觉的无人车目标体积自动测量方法，实现了无人车与目标体积测量功能的结合。通过在LeGO-LOAM算法中加入点云畸变补偿，相较于原始LeGO-LOAM算法，无人车在高速情况下的构图精度得到提升；通过将激光雷达与视觉进行深度融合，实现了目标的自动识别与全局定位；通过基于平面拟合的地面分割与欧式聚类，实现了目标点云轮廓的实时获取；通过设计一种基于切片法的不规则物体体积测量方法，实现了无人车在运动情况下对目标体积的自动估计。最终，分别通过Gazebo仿真和实际试验验证了算法的有效性。试验结果表明，所提算法在无人车运动的情况下对静态目标物的实时体积测量精度优于3%，具有较好的工程应用价值。

Automatic measurement of target volume for unmanned vehicle based on lidar/vision

In recent years, target sensing based on lidar and vision has been widely used in unmanned systems. Volume measurement of objects can play an extremely important role in many application scenarios. However, there is still no large amount of research on volume measurement for recognition and perception of objects. An automatic measurement method of target volume based on lidar/vision for unmanned vehicle is proposed for the first time. This method realizes the combination of the functions of unmanned vehicle and target volume measurement. By adding point cloud distortion compensation to LeGO-LOAM algorithm, the composition accuracy of unmanned vehicle at high speed is improved compared with the original LeGO-LOAM algorithm. Through the deep fusion of lidar and vision, the automatic recognition and global positioning of the target are realized. Through the ground segmentation and Euclidean clustering based on plane fitting, the contour of the target point cloud is obtained in real time. An irregular object volume measurement method based on slice method is designed to realize the automatic estimation of target volume under the condition of moving unmanned vehicle. Finally, the effectiveness of the algorithm is verified by Gazebo simulation and real experiment respectively. The experimental results show that the real-time volume measurement accuracy of the proposed algorithm is better than 3% when the unmanned vehicle is moving, which has good engineering application value.