Deep learning for conifer/deciduous classification of airborne LiDAR 3D point clouds representing individual trees

Abstract

The purpose of this study was to investigate the use of deep learning for coniferous/deciduous classification of individual trees from airborne LiDAR data. To enable efficient processing by a deep convolutional neural network (CNN), we designed two discrete representations using leaf-off and leaf-on LiDAR data: a digital surface model with four channels (DSM×4) and a set of four 2D views (4×2D). A training dataset of labeled tree crowns was generated via segmentation of tree crowns, followed by coregistration with field data. Potential mislabels due to GPS error or tree leaning were corrected using a statistical ensemble filtering procedure. Because the training data was heavily unbalanced (~8% conifers), we trained an ensemble of CNNs on random balanced sub-samples of augmented data (180 rotational variations per instance). The 4×2D representation yielded similar classification accuracies to the DSM×4 representation (~82% coniferous and ~90% deciduous) while converging faster. The data augmentation improved the classification accuracies, but more real training instances (especially coniferous) likely results in much stronger improvements. Leaf-off LiDAR data were the primary source of useful information, which is likely due to the perennial nature of coniferous foliage. LiDAR intensity values also proved to be useful, but normalization yielded no significant improvements. As we observed, large training data may compensate for the lack of a subset of important domain data. Lastly, the classification accuracies of overstory trees (~90%) were more balanced than those of understory trees (~90% deciduous and ~65% coniferous), which is likely due to the incomplete capture of understory tree crowns via airborne LiDAR. Automatic derivation of optimal features via deep learning provide the opportunity for remarkable improvements in prediction tasks where captured data are not friendly to human visual systemlikely yielding sub-optimal human-designed features.

Semantic Labeling of Als Point Clouds for Tree Species Mapping Using the Deep Neural Network Pointnet++

“…Presumably, one reason is the lack of large training datasets. Just recently, Hamraz et al (2018) use a CNN along with 2D images generated from ALS point clouds to classify coniferous and deciduous trees with 92% and 86% accuracy, respectively. So far, the direct usage of 3D data in DNNs for 3D vegetation mapping is more uncommon.…”

Semantic Segmentation of Manmade Landscape Structures in Digital Terrain Models

“…They are used for landuse classification in remote sensing images (Castelluccio et al, 2015), feature extraction and classification of hyperspectral images (Chen et al, 2016, Chen et al, 2014, classification and segmentation of ALS data (Yang et al, 2018), ground and multi-class land cover classification (Rizaldy et al, 2018), tree classification (Hamraz et al, 2018), ground point extraction and classification (Hu , Yuan, 2016), pointwise and object-based classification , classification of archaeological objects (Kazimi et al, 2018), forest tree detection and segmentation in ALS data (Windrim , Bryson, 2018), and semantic segmentation in ALS data , among others. While there are many applications of deep learning and CNNs for ALS data, there is still room for improvements.…”

Individual Tree-Crown Detection and Species Classification in Very High-Resolution Remote Sensing Imagery Using a Deep Learning Ensemble Model

“…For instance, the authors of [17,18] implemented deep learning fusion algorithms of hyperspectral/WorldView imagery and LiDAR data for tree species mapping with significantly better results than traditional methods or older deep learning algorithms; the authors of [19] implemented a cascade neural network for single tree detection in high-resolution remote sensing imagery; UAV images were used by the authors of [20], who implemented deep learning to detect damaged fir trees in forests, and by the authors of [21], who used a transfer learning technique for tree detection in RGB imagery. Understory trees have also been mapped with high accuracy using deep learning, as shown in [22], whose authors trained a convolutional neural network (CNN) on an airborne LiDAR.…”