鈴鹿山脈地域における空撮点群情報からのモミの個体分布抽出2026 · Temperate forests in Japan are generally divided into cool-temperate and warm-temperate zones, but the “intermediate temperate forests” located between them have not yet been clearly defined in terms of classification criteria or vegetation characteristics. Abies firma, an evergreen conifer adapted to cool-temperate climates, is regarded as an important indicator species for these transitional zones. The objective of this study was to identify the individual distribution of Abies firma within natural forests by combining UAV aerial imagery with LiDAR-derived canopy height data, applying machine learning and crown-based classification methods. UAV surveys were conducted in three different seasons, and automatic detection of Abies individuals was performed. The highest detection accuracy was obtained from imagery captured in April, when forest canopy colors showed clear seasonal variation and shadows were minimal. In this case, recall reached 71% and precision 86%. Furthermore, analysis of the relationship between detected Abies distribution and topographic factors revealed that shorter individuals tended to occur more frequently in gently sloping flat areas. These findings suggest that intermediate temperate forests dominated by Abies provide favorable conditions for accurate detection, and they demonstrate the effectiveness of remote sensing approaches for automated tree species identification.
空撮点群情報のオブジェクト指向型樹冠分類と機械学習の併用による友ヶ島植生量把握2025 · In order to promote sustainable forest management, it is essential to elaborate effective measures to detect forest species and spatial structures. Recent rapid development of remote sensing technologies including drone and LiDAR could allow such a new forest measurement. In this study, we used Tomogashima Islands in central Japan and attempted to detect individual tree canopies from drone imagery and LiDAR dataset covering the whole island through dual uses of object-based image analysis and machine learning methods. As the results, we were able to detect 4378 tree crowns of ubame oak (<i>Quercus phillyreoides</i>), dominant tree species in the island, among total 45129 tree crowns. Although we still need further field validations and various parametrical applications to other sites, our methods could have potential to classify tree crowns in the mixed broadleaf forest unlike existing methods mainly for conifer plantations and similar artificial forest types.
