基于机载LiDAR和Mean Shift算法提取森林密度

为获取森林密度信息,利用Mean Shift算法对森林点云进行单木分割提取森林密度信息.首先,以点云三维坐标和法向量作为特征向量,利用统计分析方法选择合适带宽及阈值,采用Mean Shift算法对点云进行初始分割;其次,对分割后的点云进行分析,加入灌木、杂草等过滤条件,得到树冠点云;然后,对树冠点云再次进行Mean Shift分割,并对每类树冠点云进行统计,以稳态点为粗略位置标记计算森林密度;最后,与地面实测数据进行验证.地面数据验证结果表明,平均计算精度达到90.0%以上,可满足林业应用需求;通过与分水岭法进行对比发现, Mean Shift方法获得的精度为92.5%,比分水岭方法70.0%高出22.5%,且避免了分水岭方法导致的过分割现象. 

Extraction of Forest Density Based on Airborne LiDAR and Mean Shift Algorithms

To extract forest density, Mean Shift segmentation algorithm is explored for point clouds processing. The first step is to use the Mean Shift algorithm for the point clouds initial segmentation depended on feature vectors, bandwidth and thresholds. Feature vectors consist of three-dimensional coordinates and normal vector of the point clouds and bandwidth and thresholds are determined by statistical analysis method. The second step is to obtain canopy point clouds through the analysis of the segmented point clouds added with filtering conditions such as shrubs and weeds. The third step is to use Mean Shift algorithm to calculate forest density. With the statistical analysis of segmented the canopy point clouds in each category, the steady state points are marked as the rough locations for individual trees. Finally, the results are verified by field measured data. It is found that the average accuracy of this method could be more than 90.0%, which meets the requirements of forestry industry. In comparison with the Watershed method, the accuracy of the Mean Shift method is 92.5%, which is higher than Watershed method with 70.0% accuracy, and it can avoid the over segmentation in the Watershed method.