基于无人机图像分析的树木胸径预测

发布时间：2018-07-02 18:21

  本文选题：无人机 + 图像分析　； 参考：《农业工程学报》2017年21期

【摘要】：树木胸径是林木资产评估中的重要参数,该文利用图像分析技术预测树木胸径可为资产评估提供参考。以银杏和法国梧桐为试验树种,通过拟合无人机正射图像中的单株树木树冠面积与胸径的关系预测树木胸径值。首先利用二型模糊聚类方法对无人机采集的纯林样地正射图像中的单株树冠进行分割,获取树冠像素面积,然后利用地面参照物计算出树冠的实际面积,并与测量的胸径值进行拟合,得出树冠面积与胸径的函数关系,林区中其他树木胸径值可基于该函数关系和其树冠面积计算得出。试验结果显示无人机正射图像中的银杏及法桐树冠面积与胸径均呈对数关系,且该文计算所得的银杏1.2 m处的胸径与实际胸径之间的平均误差约为0.31 cm,法桐1 m处的胸径与实际胸径之间的平均误差为0.27 cm,均在行业允许的1 cm误差范围内,该文提出的基于无人机正射图像分析技术预测树木胸径较为准确,可为中小尺度林地资产评估提供参考。
[Abstract]:Tree DBH is an important parameter in the evaluation of forest assets. In this paper, the prediction of DBH by image analysis can be used as a reference for the evaluation of forest assets. Using ginkgo biloba and sycamore as experimental tree species, the tree DBH values were predicted by fitting the relationship between tree crown area and DBH in forward shot images of UAV. In this paper, we first segment the single tree crown in the orthophoto image of pure forest sample collected by UAV by using type 2 fuzzy clustering method, and obtain the pixel area of tree crown, and then calculate the actual area of tree crown by using ground reference material. The relationship between crown area and DBH can be obtained by fitting the measured DBH value, and other tree DBH values in forest area can be calculated on the basis of this function and its crown area. The experimental results showed that the crown area of Ginkgo and Fructus were logarithmic with the diameter of DBH in the forward projection images of UAV. The average error between DBH and DBH of Ginkgo biloba at 1.2 m and actual DBH is about 0.31 cm, and the average error between DBH and actual DBH at Faton 1 m is 0.27 cm, which is within the allowable error range of 1 cm in industry. The forward projection image analysis technique based on UAV proposed in this paper is more accurate in predicting the DBH of trees, which can be used as a reference for the evaluation of forest assets in small and medium scale.
【作者单位】： 北京林业大学信息学院;
【基金】：中央高校基本科研业务费专项资金(No.2015ZCQ—XX)
【分类号】：S771.5;TP391.41
，

本文编号：2090624