International Journal of New Developments in Engineering and Society, 2024, 8(2); doi: 10.25236/IJNDES.2024.080217.
Lulu Liu, Leifang Xie, Peng Zhang, Yong Wang, Han Tian
The 27th Research Institute of China Electronics Technology Group Corporation, Zhengzhou, 450047, China
Aiming at the recognition of dead standing trees in UAV RGB images in forest areas, this paper proposes a dead standing tree detection method based on RetinaNet target detection network. Using UAV images for RetinaNet target detection network model training, and this paper compares multiple feature extraction networks, the results show that ResNet-152-FPN is the Best, using ResNet-152-FPN-based RetinaNet for target detection on dead standing trees in forest areas, the average accuracy of dead standing tree recognition reaches 81.6%. It proves the feasibility of RetinaNet target detection network on dead standing wood recognition, and provides strong support for dead standing wood census recognition in forest areas.
UAV RGB images, RetinaNet, target detection, dead standing trees
Lulu Liu, Leifang Xie, Peng Zhang, Yong Wang, Han Tian. Method of Dead Standing Tree Detection Based on RetinaNet Object Detection Network. International Journal of New Developments in Engineering and Society (2024) Vol.8, Issue 2: 110-114. https://doi.org/10.25236/IJNDES.2024.080217.
[1] Oettel J, Lapin K, Kindermann G, et al. Patterns and drivers of deadwood volume and composition in different forest types of the Austrian natural forest reserves[J]. Forest Ecology and Management, 2020, 463(05): 118016.
[2] Franklin J, Shugart H, Harmon M. Tree Death as an Ecological Process[J]. Bioscience, 1987, 37(8): 550-556.
[3] Lu Zhijun, Liu Fuling, Wu Hao, et al. Species composition, size class, and distribution pattern of dead standing trees in the evergreen and deciduous broad-leaved mixed forest of Badagongshan Mountain[J]. Biodiversity Science, 2015, 23(02): 167-173.
[4] Guo Yili, Wang Bin, Xiang Wusheng, et al. Spatial pattern and habitat association analysis of dead standing trees in the seasonal rainforest of southwestern Guangxi[J]. Guihaia, 2016, 36(02): 154-161.
[5] An Yun, Ding Guodong, Gao Guanglei, et al. Quantity characteristics and distribution pattern of dead standing trees in natural secondary forests in the North China Loess Plateau[J]. Bulletin of Soil and Water Conservation, 2012, 32(04): 246-250.
[6] Heurich, Marco, Nielsen, et al. Automatic Mapping of Standing Dead Trees after an Insect Outbreak Using the Window Independent Context Segmentation Method[J]. Journal of Forestry, 2014,112(6):564-571.
[7] Polewski P, Shelton J, Yao W, et al. Segmentation of single standing dead trees in high-resolution aerial imagery with generative adversarial network-based shape priors[J]. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2020,XLIII-B2-2020:717-723.
[8] Briechle S, Krzystek P, Vosselman G. Silvi-Net – A dual-CNN approach for combined classification of tree species and standing dead trees from remote sensing data[J]. International Journal of Applied Earth Observation and Geoinformation, 2021,98:102292.
[9] Lin T Y, Goyal P, Girshick R, et al. Focal Loss for Dense Object Detection[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2018,PP(99):2999-3007.
[10] He K, Zhang X, Ren S, et al. Deep Residual Learning for Image Recognition: IEEE Conference on Computer Vision & Pattern Recognition, Las Vegas, NV, USA, 2016[C].
[11] Lin T Y, Dollár P, Girshick R, et al. Feature Pyramid Networks for Object Detection: IEEE Conference on Computer Vision & Pattern Recognition, Las Vegas, NV, USA, 2016[C].