Identification Of Tree Crowns Research Articles

Tree Recognition and Crown Width Extraction Based on Novel Faster-RCNN in a Dense Loblolly Pine Environment

Tree crown width relates directly to wood quality and tree growth. The traditional method used to measure crown width is labor-intensive and time-consuming. Pairing imagery taken by an unmanned aerial vehicle (UAV) with a deep learning algorithm such as a faster region-based convolutional neural network (Faster-RCNN) has the potential to be an alternative to the traditional method. In this study, Faster-RCNN outperformed single-shot multibox detector (SSD) for crown detection in a young loblolly pine stand but performed poorly in a dense, mature loblolly pine stand. This paper proposes a novel Faster-RCNN algorithm for tree crown identification and crown width extraction in a forest stand environment with high-density loblolly pine forests. The new algorithm uses Residual Network 101 (ResNet101) and a feature pyramid network (FPN) to build an FPN_ResNet101 structure, improving the capability to model shallow location feature extraction. The algorithm was applied to images from a mature loblolly pine plot in eastern Texas, USA. The results show that the accuracy of crown recognition and crown width measurement using the FPN_ResNet101 structure as the backbone network in Faster-RCNN (FPN_Faster-RCNN_ResNet101) was high, being 95.26% and 0.95, respectively, which was 4.90% and 0.27 higher than when using Faster-RCNN with ResNet101 as the backbone network (Faster-RCNN_ResNet101). The results fully confirm the effectiveness of the proposed algorithm.

Automated detection of individual juniper tree location and forest cover changes using Google Earth Engine

Tree detection is the first step in the appraisal of a forest, especially when the focus is monitoring the growth of tree canopy. The acquisition of annual very high-resolution aerial images by the National Agriculture Imagery Program (NAIP) and their accessibility through Google Earth Engine (GEE) supports the delineation of tree canopies and change over time in a cost and time-effective manner. The objectives of this study are to develop an automated method to detect the crowns of individual western Juniper (Juniperus occidentalis) trees and to assess the change of forest cover from multispectral 1-meter resolution NAIP images collected from 2009 to 2016 in Oregon, USA. The Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and Ratio Vegetation Index (RVI) were calculated from the NAIP images, in addition to the red-green-blue-near infrared bands. To identify the most suitable approach for individual tree crown identification, we created two training datasets: one considering yearly images separately and one merging all images, irrespective of the year. We segmented individual tree crowns using a random forest algorithm implemented in GEE and seven rasters, namely the reflectance of four spectral bands as recorded by the NAIP images (i.e., the red-green-blue-near infrared) and three calculated indices (i.e., NDVI, NDWI, and RVI). We compared the estimated location of the trees, computed as the centroid of the crown, with the visually identified treetops, which were considered as validation locations. We found that tree location errors were smaller when years were analyzed individually than by merging the years. Measurements of completeness (74%), correctness (94%), and mean accuracy detection (82 %) show promising performance of the random forest algorithm in crown delineation, considering that only four original input bands were used for crown segmentation. The change in the calculated crown area for western juniper follows a sinusoidal curve, with a decrease from 2011 to 2012 and an increase from 2012 to 2014. The proposed approach has the potential to estimate individual tree locations and forest cover area dynamics at broad spatial scales using regularly collected airborne imagery with easy-to-implement methods.

Automated detection of individual Juniper tree location and forest cover changes using Google Earth Engine

Tree detection is the first step in the appraisal of a forest, especially when the focus is monitoring the growth of tree canopy. The acquisition of annual very high-resolution aerial images by the National Agriculture Imagery Program (NAIP) and their accessibility through Google Earth Engine (GEE) supports the delineation of tree canopies and change over time in a cost and time-effective manner. The objectives of this study are to develop an automated method to detect the crowns of individual western Juniper (Juniperus occidentalis) trees and to assess the change of forest cover from multispectral 1-meter resolution NAIP images collected from 2009 to 2016, in Oregon, USA. The Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and Ratio Vegetation Index (RVI), were calculated from the NAIP images, in addition to the red-green-blue-near infrared bands. To identify the most suitable approach for individual tree crown identification, we created two training datasets: one considering yearly images separately and one merging all images, irrespective of the year. We segmented individual tree crowns using a random forest algorithm implemented in GEE and seven rasters, namely the reflectance of four spectral bands as recorded by the NAIP images (i.e., the red-green-blue-near infrared) and three calculated indices (i.e., NDVI, NDWI, and RVI). We compared the estimated location of the trees, computed as the centroid of the crown, with the visually identified treetops, which were considered as validation locations. We found that tree location errors were smaller when years were analyzed individually than by merging the years. Measurements of completeness (74%), correctness (94%), and mean accuracy detection (82 %) show promising performance of the random forest algorithm in crown delineation, considering that only four original input bands were used for crown segmentation. The change in the calculated crown area for western juniper follows a sinusoidal curve, with a decrease from 2011 to 2012 and an increase from 2012 to 2014. The proposed approach has the potential to estimate individual tree locations and forest cover area dynamics at broad spatial scales using regularly collected airborne imagery with easy-to-implement methods.

Bridging the genotype-phenotype gap for a Mediterranean pine by semi-automatic crown identification and multispectral imagery.

Progress in high-throughput phenotyping and genomics provides the potential to understand the genetic basis of plant functional differentiation. We developed a semi-automatic methodology based on unmanned aerial vehicle (UAV) imagery for deriving tree-level phenotypes followed by genome-wide association study (GWAS). An RGB-based point cloud was used for tree crown identification in a common garden of Pinus halepensis in Spain. Crowns were combined with multispectral and thermal orthomosaics to retrieve growth traits, vegetation indices and canopy temperature. Thereafter, GWAS was performed to analyse the association between phenotypes and genomic variation at 235 single nucleotide polymorphisms (SNPs). Growth traits were associated with 12 SNPs involved in cellulose and carbohydrate metabolism. Indices related to transpiration and leaf water content were associated with six SNPs involved in stomata dynamics. Indices related to leaf pigments and leaf area were associated with 11 SNPs involved in signalling and peroxisome metabolism. About 16-20% of trait variance was explained by combinations of several SNPs, indicating polygenic control of morpho-physiological traits. Despite a limited availability of markers and individuals, this study is provides a successful proof-of-concept for the combination of high-throughput UAV-based phenotyping with cost-effective genotyping to disentangle the genetic architecture of phenotypic variation in a widespread conifer.

INTEGRATION OF TREE DATABASE DERIVED FROM SATELLITE IMAGERY AND LIDAR POINT CLOUD DATA

Abstract. 3D tree database provides essential information of tree species abundance, spatial distribution and tree height for forest mapping, sustainable urban planning and 3D city modelling. Fusion of passive optical satellite imagery and active Lidar data can potentially be exploited for operational forest inventory. However, such fusion requires very high geometric accuracy for both data sets. This paper proposes an approach for 3D tree information extracted from passive and active data integrating into existing tree database by effectively using geometric information of satellite camera model and laser scanner scanning geometry. The paper also presents the individual methods for tree crown identification and delineation from satellite images and lidar point cloud data respectively, the geometric correction of tree position from tree top to tree base. The ground truth accuracy assessment for the tree extracted is also present.

Delineation of Tree Crowns and Tree Species Classification From Full-Waveform Airborne Laser Scanning Data Using 3-D Ellipsoidal Clustering

Individual tree crowns can be delineated from dense airborne laser scanning (ALS) data and their species can be classified from the spatial distribution and other variables derived from the ALS data within each tree crown. This study reports a new clustering approach to delineate tree crowns in three dimensions (3-D) based on ellipsoidal tree crown models (i.e., ellipsoidal clustering). An important feature of this approach is the aim to derive information also about the understory vegetation. The tree crowns are delineated from echoes derived from full-waveform (fwf) ALS data as well as discrete return ALS data with first and last returns. The ellipsoidal clustering led to an improvement in the identification of tree crowns. FwfALS data offer the possibility to derive also the echo width and the amplitude in addition to the 3-D coordinates of each echo. In this study, tree species are classified from variables describing the fwf (i.e., the mean and standard deviation of the echo amplitude, echo width, and total number of echoes per pulse) and the spatial distribution of the clusters for pine, spruce, birch, oak, alder, and other species. Supervised classification is done for 68 field plots with leave-one-out cross-validation for one field plot at a time. The total accuracy was 71% when using both fwf and spatial variables, 60% when using only spatial variables, and 53% when using discrete return data. The improvement was greatest for discriminating pine and spruce as well as pine and birch.

Markov-random-field-based super-resolution mapping for identification of urban trees in VHR images

Identification of tree crowns from remote sensing requires detailed spectral information and submeter spatial resolution imagery. Traditional pixel-based classification techniques do not fully exploit the spatial and spectral characteristics of remote sensing datasets. We propose a contextual and probabilistic method for detection of tree crowns in urban areas using a Markov random field based super resolution mapping (SRM) approach in very high resolution images. Our method defines an objective energy function in terms of the conditional probabilities of panchromatic and multispectral images and it locally optimizes the labeling of tree crown pixels. Energy and model parameter values are estimated from multiple implementations of SRM in tuning areas and the method is applied in QuickBird images to produce a 0.6 m tree crown map in a city of The Netherlands. The SRM output shows an identification rate of 66% and commission and omission errors in small trees and shrub areas. The method outperforms tree crown identification results obtained with maximum likelihood, support vector machines and SRM at nominal resolution (2.4 m) approaches.

Estimating plot-level tree heights with lidar: local filtering with a canopy-height based variable window size

In recent years, the use of airborne lidar technology to measure forest biophysical characteristics has been rapidly increasing. This paper discusses processing algorithms for deriving the terrain model and estimating tree heights by using a multiple return, high–density, small-footprint lidar data set. The lidar data were acquired over deciduous, coniferous, and mixed stands of varying age classes and settings typical of the southeastern US. The specific objectives were: (1) to develop and test algorithms to estimate plot level tree height using lidar data, and (2) to investigate how ground measurements can help in the processing phase of lidar data for tree height assessment. The study area is located in the Piedmont physiographic province of Virginia, USA and includes a portion of the Appomattox-Buckingham State Forest (37°25′N, 78°41′W). Two lidar processing algorithms are discussed—the first based on single tree crown identification using a variable window size for local filtering, and the second based on the height of all laser pulses within the area covered by the ground truth data. Height estimates resulted from processing lidar data with both algorithms were compared to field measurements obtained with a plot design following the USDA Forest Service Forest Inventory and Analysis (FIA) field data layout. Linear regression was used to develop equations relating lidar-estimated parameters with field inventories for each of the FIA plots. As expected, the maximum height on each plot was predicted with the highest accuracy ( R 2 values of 85 and 90%, for the first and the second algorithm, respectively). The variable window size algorithm performed better for predicting heights of dominant and co-dominant trees ( R 2 values 84–85%), with a diameter at breast height (dbh) larger than 12.7 cm (5 in), when compared with the algorithm based on all laser heights ( R 2 values 57–73%). The use of field-based height thresholds when processing lidar data did not bring significant gains in explaining the total variation associated with tree height. The technique of local filtering with a variable window size considers fundamental forest biometrics relationships and overall proved to give better results than the technique of all laser shots.