無人機可見光遙感影像的空間分辨率對毛竹立竹度識別精度的影響1)

陸雪婷 鄧洋波 宋賢芬 嚴夏帆 余坤勇 劉健

(福建農林大學，福州，350002) (福州市林業局) (福建農林大學)

In order to explore the applicability of UAV visible image in the extraction of Phyllostachys edulis standing degree, the P. edulis forest in Tianbaoyan National Nature Reserve, Yongan City, Sanming City, Fujian Province, was taken as the research object, and the visible image under different spatial resolutions (0.25, 0.5 and 0.75 m) was obtained by resampling method. Object-oriented multi-scale segmentation was used to determine the optimal segmentation scale under each resolution, and random forest classification was used to compare the extraction accuracy of P. edulis stand density under different spatial resolutions. The results show that: the best segmentation scale for the segmentation scale to select different under different image spatial resolution of 0.25, 0.50 and 0.75 m are 20, 9 and 8, P. edulis stand density extract accuracy are 80.97%, 81.29% and 77.82%, and Kappa coefficients are 0.806 0, 0.863 3 and 0.817 1, respectively. When the spatial resolution of image is 0.50 m, the overall classification accuracy and Kappa coefficient are the highest. Therefore, the appropriate spatial resolution of image extraction of P. edulis stand density is 0.5 m, and the optimal segmentation scale is 9.

毛竹立竹度也叫竹林密度，指每公頃毛竹林地中毛竹的立木株數。也是毛竹林群體結構和生產力指標最重要的數量特征，同時也是廣大竹農對毛竹生長發育期間干預竹林結構的重要因子[1-2]。目前關于毛竹立竹度的監測主要采用人工調查和多光譜影像提取。采用人工調查不但需要更多人力，而且由于地理因素導致人員可及度受到限制；利用多光譜影像提取毛竹立竹度，受到影像波段的復雜性和影像獲取的成本的限制。隨著無人機遙感監測技術的不斷發展，為利用可見光遙感監測調查立竹度提供了技術支持，也為調整竹林的密度，改善毛竹林分結構，提高毛竹林的生產力提供了手段，對我國竹林經營具有重要的實際意義[3]。

利用遙感手段提取立竹度的關鍵是對影像的有效分割和準確分類。目前國內外學者在基于圖像分割技術的林木冠層特征提取方面已經取得了很大的進展，孫曉艷等[4]利用SPOT衛星數據同時結合紋理信息，通過面向對象的方法構造多尺度、多層次的結構來實現竹林信息的提取；采用面向對象的多尺度分割方法得到分割后的目標，不僅可以有效地獲得識別單元的幾何特征，而且可以提取各種地物的信息[5]；李越帥等[6]提出了深度學習與分水嶺分割相結合的處理方法，對密集胡楊樹冠的精確分割和胡楊個體信息的提取；……