基于无人机遥感的亚热带森林林分株数提取

Counting Tree Number in Subtropical Forest Districts based on UAV Remote Sensing Images

株数是描述林分密度的指标,株数提取对于森林资源信息调查具有重要的意义。以福建省将乐县林场为研究区,探讨了局域最大值法和多尺度分割算法在森林株数提取中的适用性。首先,利用无人机遥感获取优于10cm分辨率影像,经过处理获得正射影像。在此基础上,分别选取针叶林、阔叶林20个样地区域;其次,使用局域最大值法和多尺度分割算法分别统计两种方法提取的样地株数;最后,根据两种算法提取的株数与目视解译统计的株数进行对比分析。结果表明:两种算法提取样地株数的总体精度在90%左右。其中局域最大值法在针叶林样地株数的提取精度明显优于阔叶林样地的株数提取精度,多尺度分割算法在阔叶林样地的株数提取精度明显优于局域最大值法的株数提取精度。研究认为,局域最大值法较适用于无人机遥感影像上针叶林样地株数的快速提取,而多尺度分割算法较适用于阔叶林样地株数的准确提取。

Tree number is the index that describes the stand density,and extracting tree number in districts is important for researching forest reserve information.This essay investigates Jiangle forest farm in Fujian to study the applicability of maximum algorithm and multiscale segmentation algorithm in extracting tree number.The first step in this dissertation is to use ebee unmanned aerial vehicle remote sensing with fixed wings to get image whose resolution is more than 10cm,and gets orthoimage after processing.Based on it, this essay uses 20 area samples including coniferous forest and broad-leaved forest. The second, the tree number of samples were extracted from maximum algorithm and multiscale segmentation algorithm;Lastly,this essay uses the tree number abstracted fromthe two algorithm and the tree number from visual interpretation statistics to precision analysis.The results show that the tree number of samples was extracted by the two algorithm in the overall accuracy of about 90%.In the coniferous forest plots the tree number of extraction accuracy by local maximum algorithms is better than the broad-leaved forest plots the tree number of extraction accuracy.In the broad-leaved forest plots the tree number of extraction accuracy by multi- scale segmentation algorithm is better than the tree number of extraction accuracy by local maximum algorithms.Therefore,this research argues that local maximum value algorithm is more suitable for unmanned aerial vehicle （uav） remote sensing image on coniferous sample tree number of rapid extraction,the multi- scale segmentation algorithm is more suitable for unmanned aerial vehicle （uav） remote sensing image on broad-leaved sample tree number of accurate extraction.