基于SFM算法的单木结构参数提取研究

Extraction of the Parameters of Single Tree Structure Based on SFM Algorithm

针对传统森林资源调查方法获取单木结构参数效率低和成本高的问题,提出一种基于SFM算法的单木结构参数快速提取方法。以哈尔滨市城市林业示范基地树木为研究对象,利用SFM算法获得单木照片的三维点云,并利用点云数据处理软件对获得的点云数据进行单木结构参数提取,最后与实测参数进行对比分析。结果表明:1)分别利用SIFT算法、SURF算法以及ORB算法对相机校检后的树木照片进行特征点提取匹配,特征点正确匹配个数分别为23、145以及25,相应的耗时分别为18.56、16.04、1.58 s;2)利用SFM算法能获得树木照片的稀疏点云和稠密点云,平均每棵树木点云量为80万个;3)基于点云数据提取单木结构参数的胸径、树高及冠幅的平均绝对误差分别为1.79 cm、0.77 m及0.79 m;胸径、树高、冠幅的提取值与实测值相关系数均>0.94。

Aiming at solving the problems of low efficiency and high cost of single tree structure parameter extraction based on the traditional forest resource survey method,a fast extraction method based on structure from motion(SFM)algorithm was proposed.Trees in Harbin Urban Forestry Demonstration Base were used as the research objects,3D point clouds of single tree photos using SFM algorithm were obtained.The point cloud data processing software was used to extract the single tree structure parameters from the obtained point cloud data,and which were finally compared with the measured parameters.The results showed that 1)three algorithms,i.e.,SIFT,SURF and ORB were used to extract and match the feature points on the camera calibrated tree photos,respectively.The numbers of feature points matching of the three algorithms were 23,145 and 25,respectively.The corresponding time consumptions were 18.56,16.04 and 1.58's respectively.2)The sparse point clouds and dense point clouds of the tree photos could be obtained by using SFM algorithm,and the average amount of cloud data per tree was 800000.3)The average absolute errors of DBH,tree height and tree crown of the single tree structure parameters based on point cloud data were 1.79 cm,0.77 m and 0.79 m,respectively;The correlation coefficient between the extracted values of the DBH,tree height and crown width and the measured values were all above 0.94.