基于无人机LiDAR的天然林与人工林林隙提取

为研究主动遥感进行森林地物分类和林隙提取的效果,分别在天然林和人工林中比较了无人机激光雷达(Li DAR)数据的阈值法、逐像元法和面向对象法3种方法的分类精度和适用性。选取天然林(黑龙江省哈尔滨市帽儿山林场)和人工林(内蒙古自治区赤峰市旺业甸林场)两处试验区,应用阈值法、逐像元法和面向对象法3种方法,对两个试验区采集的无人机Li DAR数据进行林隙、非林隙、其他类型划分。研究结果表明,面向对象法在天然林和人工林试验区中的分类精度和Kappa系数均最高,天然林为82. 43%、0. 73,人工林为91. 74%、0. 88;逐像元法次之,天然林为76. 62%、0. 64,人工林为78. 68%、0. 68;阈值法的分类精度和Kappa系数差异较大,在天然林中的精度极低,为50. 54%、0. 27,人工林的精度较高,为79. 12%、0. 69。面向对象法和逐像元法在天然林和人工林普遍适用,均可以达到理想的分类精度和Kappa系数。阈值法在天然林的精度较低,更适合于人工林的分类,即林分高度趋于一致,且建筑、道路等其他类型干扰较少的区域。天然林的最佳分类方法为面向对象法,人工林的最佳分类方法为阈值法。

Extraction of Forest Gaps in Natural Forest and Man-made Forest Based on UAV LiDAR

Aiming to explore results of classification and extraction of forest gaps in forest based on active remote sensing, the classification accuracy and applicability of the threshold method were compared, including the pixel oriented method and the object oriented method in natural forest and man-made forest based on unmanned aerial vehicle (UAV) light detection and ranging (LiDAR). Maoershan Forest Farm in Harbin, Heilongjiang Province and Wangyedian Forest Farm in Chifeng, Inner Mongolia Autonomous Region were selected as the natural forest experimental site and the man made forest experimental site respectively. The threshold method, the pixel-oriented method and the object-oriented method were applied to classify the two experimental sites into three classes based on UAV LiDAR acquired, which were forest gap, non forest gap and others. The research results indicated that the object-oriented method produced the highest classification accuracy and Kappa coefficient in both experimental sites, which were 82.43% and 0.73 in natural forest and 91.74% and 0.88 in man-made forest, respectively. The accuracy and Kappa coefficient of the pixel-oriented method was lower than that of the object oriented method, which was 76.62% and 0.64 in natural forest and 78.68% and 0.68 in man-made forest. The accuracy and Kappa coefficient of the threshold method had larger difference. It produced the lowest accuracy in natural forest (50.54% and 0.27) and the higher accuracy in man-made forest (79.12% and 0.69). The object oriented method and the pixel-oriented method were the methods with general applicability in classification of natural forest and man-made forest and could produce ideal accuracy and Kappa coefficient. The threshold method produced lower accuracy in natural forest and was more suitable for classification of man-made forest, forest height of which was similar and where others such as buildings and roads were rare. The best classification method of natural forest was the object-oriented method and the best of man-made forest was the threshold method. The research results provided method reference and technology support for extraction of forest gaps in natural forest and man-made forest.