基于系统聚类的林地内采育目标识别与分类

为了避免多功能林木联合采育装备的采育工作装置的误操作,利用激光扫描仪获取林地内采育作业环境内的目标数据,运用基于腐蚀和聚类原理的滤波算法过滤原始扫描数据的背景噪声,获取扫描目标的轮廓数据.并假设所有目标为标准圆,利用最小二乘法拟合目标半径,平均误差小于4.29 mm.采用一种基于多元统计分析的系统聚类方法对拟合结果数据进行分类,区分采育目标立木和大型障碍物.试验结果表明,采用以上方法可以有效区分林地内的拟合半径小于384 mm采育目标立木与拟合半径大干774 mm的大犁障碍物.

Identification and classification of scanned target in forest based on hierarchical cluster

In order to avoid the incorrect operation of the harvesting head of forest harvester, the range data of the target trees were collected with a laser scanner. The backgrounds of the scanned data were filtered with the filtering algorithm based on principles of corrosion and clustering. Then, the outline of the scanned target was drawn. On the assumption that all cross sections of the target trees were standard circulars, then the radiuses of them were calculated with the least square method and the mean error of which was less than 4.29 mm. At last, a kind of clustering method based on multivariate statistical analysis was used to classify the calculated results, so the target trees and the large obstacles could be classified based on hierarchical cluster. The final experimental result showed that the method could distinguish the target trees with radius smaller than 384 mm and the large obstacles with calculated radius larger than 774 mm effectively.