In modeling forest stand growth and yield,crown width,a measure for stand density,is among the parameters that allows for estimating stand timber volumes.However,accurately measuring tree crown size in the field,in pa...In modeling forest stand growth and yield,crown width,a measure for stand density,is among the parameters that allows for estimating stand timber volumes.However,accurately measuring tree crown size in the field,in particular for mature trees,is challenging.This study demonstrated a novel method of applying machine learning algorithms to aerial imagery acquired by an unmanned aerial vehicle(UAV)to identify tree crowns and their widths in two loblolly pine plantations in eastern Texas,USA.An ortho mosaic image derived from UAV-captured aerial photos was acquired for each plantation(a young stand before canopy closure,a mature stand with a closed canopy).For each site,the images were split into two subsets:one for training and one for validation purposes.Three widely used object detection methods in deep learning,the Faster region-based convolutional neural network(Faster R-CNN),You Only Look Once version 3(YOLOv3),and single shot detection(SSD),were applied to the training data,respectively.Each was used to train the model for performing crown recognition and crown extraction.Each model output was evaluated using an independent test data set.All three models were successful in detecting tree crowns with an accuracy greater than 93%,except the Faster R-CNN model that failed on the mature site.On the young site,the SSD model performed the best for crown extraction with a coefficient of determination(R^(2))of 0.92,followed by Faster R-CNN(0.88)and YOLOv3(0.62).As to the mature site,the SSD model achieved a R^(2)as high as 0.94,follow by YOLOv3(0.69).These deep leaning algorithms,in particular the SSD model,proved to be successfully in identifying tree crowns and estimating crown widths with satisfactory accuracy.For the purpose of forest inventory on loblolly pine plantations,using UAV-captured imagery paired with the SSD object detention application is a cost-effective alternative to traditional ground measurement.展开更多
To improve wood quality for pulpwood industries, it is important to examine not only wood density but also its components, especially tracheid characteristics. We studied genetic variations in the following tracheid t...To improve wood quality for pulpwood industries, it is important to examine not only wood density but also its components, especially tracheid characteristics. We studied genetic variations in the following tracheid traits by earlywood （EW） and latewood （LW）： tracheid length （TL）, double wall thickness （WT）, radial lumen diameter （R_D1）, tangential lumen diameter （T_D1）, radial central diameter （R_D2）, and tangential central diameter （T_D2）. We also studied the relationship with the following growth traits： diameter at breast height （DBH）, height （H）, crown breadth south-north axis （NSC）, crown breadth east-west axis （EWC）, ring width （RW）, latewood percentage （LWP）, and wood density （WD）. All sample materials were collected from a 33-year old clonal seed orchard of Pinus tabuliformis Carr. Genetic variation among clones was moderate for all tracheid traits, 9.49-26.03%. Clones significantly affected WT, R_D1, R_D2, T_D1, T_D2, and the two ratios WT/R_D1 and TL/T_D2 in EW but had no effects in LW. Clones significantly affected TL in LW but had no effects in EW. H2/C was higher in LW （0.50） than in EW （0.20） for TL, while H 2/C was higher in EW （0.27-0.46） for other tracheid traits and the two ratios （TL/T_D2 and WT/R_D1） than in EW （0.06-0.22）. WD and TL were significantly positively correlated, but WT and TL were negatively correlated both at individual and clone levels; all tracheid diameters and the four ratio values （EW_WT/ R_D1, LW WT/R_D1, EW_TL/T_D2 and LW_TL/ T_D2）, were strongly positively correlated with DBH, H, NSC, WEC and RW, and strongly negatively correlated with WD both at individual and clone levels. The most important variables for predicting WD were LW_TL, EW_WT and R_D1 in both EW and LW （r2= 0.22）. Selecting the top 10% of the clones by DBH would improve DBH growth by 12.19% （wood density was reduced by 0.14%） and produced similar responses between EW and LW for all tracheid traits： a reduction of 0.94 and 3.69% in tracheid length and increases in tracheid diameters （from 0.36 to 5.24%） and double wall thickness （0.07 and 0.87%）. The two ratios WT/R_D1 and TL/T_D2 across tissues （EW and LW） declined 0.59 and 4.56%, respectively. The decreased tracheid length and the ratio between tracheid length and diameter is disadvantageous for pulp production. The unfavorable relationship of tracheid traits with wood density indicate that multiple trait selection using optimal economic weights and optimal breeding strategies are recommended for the current longterm breeding program for P. tabuliformis.展开更多
基金supported by the Mc IntireStennis program and East Texas Pine Plantation Research Project at Stephen F.Austin State UniversityPart of the research was also supported by Zhejiang Provincial Key Science and Technology Project(2018C02013)。
文摘In modeling forest stand growth and yield,crown width,a measure for stand density,is among the parameters that allows for estimating stand timber volumes.However,accurately measuring tree crown size in the field,in particular for mature trees,is challenging.This study demonstrated a novel method of applying machine learning algorithms to aerial imagery acquired by an unmanned aerial vehicle(UAV)to identify tree crowns and their widths in two loblolly pine plantations in eastern Texas,USA.An ortho mosaic image derived from UAV-captured aerial photos was acquired for each plantation(a young stand before canopy closure,a mature stand with a closed canopy).For each site,the images were split into two subsets:one for training and one for validation purposes.Three widely used object detection methods in deep learning,the Faster region-based convolutional neural network(Faster R-CNN),You Only Look Once version 3(YOLOv3),and single shot detection(SSD),were applied to the training data,respectively.Each was used to train the model for performing crown recognition and crown extraction.Each model output was evaluated using an independent test data set.All three models were successful in detecting tree crowns with an accuracy greater than 93%,except the Faster R-CNN model that failed on the mature site.On the young site,the SSD model performed the best for crown extraction with a coefficient of determination(R^(2))of 0.92,followed by Faster R-CNN(0.88)and YOLOv3(0.62).As to the mature site,the SSD model achieved a R^(2)as high as 0.94,follow by YOLOv3(0.69).These deep leaning algorithms,in particular the SSD model,proved to be successfully in identifying tree crowns and estimating crown widths with satisfactory accuracy.For the purpose of forest inventory on loblolly pine plantations,using UAV-captured imagery paired with the SSD object detention application is a cost-effective alternative to traditional ground measurement.
基金supported by “Open Fund of State Key Laboratory of Tree Genetics and Breeding(Chinese Academy of Forestry)(Grant No.TGB2016001)”“The Lecture and Study Program for Outstanding Scholars from Home and Abroad(Grant No.CAFYBB2011007)”“Continuation project of National Natural Science Foundation of China(Grant No.CAFNSFC201601)”
文摘To improve wood quality for pulpwood industries, it is important to examine not only wood density but also its components, especially tracheid characteristics. We studied genetic variations in the following tracheid traits by earlywood （EW） and latewood （LW）： tracheid length （TL）, double wall thickness （WT）, radial lumen diameter （R_D1）, tangential lumen diameter （T_D1）, radial central diameter （R_D2）, and tangential central diameter （T_D2）. We also studied the relationship with the following growth traits： diameter at breast height （DBH）, height （H）, crown breadth south-north axis （NSC）, crown breadth east-west axis （EWC）, ring width （RW）, latewood percentage （LWP）, and wood density （WD）. All sample materials were collected from a 33-year old clonal seed orchard of Pinus tabuliformis Carr. Genetic variation among clones was moderate for all tracheid traits, 9.49-26.03%. Clones significantly affected WT, R_D1, R_D2, T_D1, T_D2, and the two ratios WT/R_D1 and TL/T_D2 in EW but had no effects in LW. Clones significantly affected TL in LW but had no effects in EW. H2/C was higher in LW （0.50） than in EW （0.20） for TL, while H 2/C was higher in EW （0.27-0.46） for other tracheid traits and the two ratios （TL/T_D2 and WT/R_D1） than in EW （0.06-0.22）. WD and TL were significantly positively correlated, but WT and TL were negatively correlated both at individual and clone levels; all tracheid diameters and the four ratio values （EW_WT/ R_D1, LW WT/R_D1, EW_TL/T_D2 and LW_TL/ T_D2）, were strongly positively correlated with DBH, H, NSC, WEC and RW, and strongly negatively correlated with WD both at individual and clone levels. The most important variables for predicting WD were LW_TL, EW_WT and R_D1 in both EW and LW （r2= 0.22）. Selecting the top 10% of the clones by DBH would improve DBH growth by 12.19% （wood density was reduced by 0.14%） and produced similar responses between EW and LW for all tracheid traits： a reduction of 0.94 and 3.69% in tracheid length and increases in tracheid diameters （from 0.36 to 5.24%） and double wall thickness （0.07 and 0.87%）. The two ratios WT/R_D1 and TL/T_D2 across tissues （EW and LW） declined 0.59 and 4.56%, respectively. The decreased tracheid length and the ratio between tracheid length and diameter is disadvantageous for pulp production. The unfavorable relationship of tracheid traits with wood density indicate that multiple trait selection using optimal economic weights and optimal breeding strategies are recommended for the current longterm breeding program for P. tabuliformis.
