Development and Validation of LiDAR Sensor Simulators Based on Parallel Raycasting

GPU Rasterization-Based 3D LiDAR Simulation for Deep Learning

High-quality data are of utmost importance for any deep-learning application. However, acquiring such data and their annotation is challenging. This paper presents a GPU-accelerated simulator that enables the generation of high-quality, perfectly labelled data for any Time-of-Flight sensor, including LiDAR. Our approach optimally exploits the 3D graphics pipeline of the GPU, significantly decreasing data generation time while preserving compatibility with all real-time rendering engines. The presented algorithms are generic and allow users to perfectly mimic the unique sampling pattern of any such sensor. To validate our simulator, two neural networks are trained for denoising and semantic segmentation. To bridge the gap between reality and simulation, a novel loss function is introduced that requires only a small set of partially annotated real data. It enables the learning of classes for which no labels are provided in the real data, hence dramatically reducing annotation efforts. With this work, we hope to provide means for alleviating the data acquisition problem that is pertinent to deep-learning applications.

Measuring tree diameter using a LiDAR-equipped smartphone: a comparison of smartphone- and caliper-based DBH

Tree diameter measurement is one of the most important stages of forest inventories to assess growing stock, aboveground biomass, and landscape restoration options, among others. This study investigates the accuracy of measuring tree diameters using a Light Detection and Ranging (LiDAR)-equipped smartphone vs. a normal caliper (reference data) and the opportunity to use low-cost smartphone-based applications in forest inventories. To estimate the diameter at breast height (DBH) of single trees, we used a smartphone with a third-party app that automatically analyzed three-dimensional (3D) point clouds. For two different measurement techniques, we compared the two measurement techniques based on DBH data from 55 Calabrian pine (Pinus brutia Ten.) and 50 oriental plane (Platanus orientalis L.) trees using the paired-sample t-test and Wilcoxon signed-rank test. Mean absolute error (MAE), mean squared error (MSE), root mean square error (RMSE), percent bias (PBIAS), and coefficient of determination (R²) were used as precision and error statistics. Statistical differences were observed between the reference and smartphone-based DBH according to the paired-sample t-test and Wilcoxon signed-rank test. The R² values obtained were determined as 0.91, 0.88, and 0.88 for Calabrian pine, oriental plane, and all tree species (105 trees), respectively. In addition to the overall accuracy performance of the comparison between reference and estimations, MAE, MSE, RMSE, and PBIAS values for the DBH of 105 tree stems were calculated as 1.56 cm, 5.42 cm, 2.33 cm, and - 5.10%, respectively. The estimation accuracies increased in regular stem forms compared with forked stems particularly observed on plane trees. Further experiments are needed to investigate the uncertainties associated with trees of different stem forms, species (coniferous or deciduous), different work environments, and different types of LiDAR and LiDAR-based app scanners.

BLAINDER-A Blender AI Add-On for Generation of Semantically Labeled Depth-Sensing Data

Common Machine-Learning (ML) approaches for scene classification require a large amount of training data. However, for classification of depth sensor data, in contrast to image data, relatively few databases are publicly available and manual generation of semantically labeled 3D point clouds is an even more time-consuming task. To simplify the training data generation process for a wide range of domains, we have developed the BLAINDER add-on package for the open-source 3D modeling software Blender, which enables a largely automated generation of semantically annotated point-cloud data in virtual 3D environments. In this paper, we focus on classical depth-sensing techniques Light Detection and Ranging (LiDAR) and Sound Navigation and Ranging (Sonar). Within the BLAINDER add-on, different depth sensors can be loaded from presets, customized sensors can be implemented and different environmental conditions (e.g., influence of rain, dust) can be simulated. The semantically labeled data can be exported to various 2D and 3D formats and are thus optimized for different ML applications and visualizations. In addition, semantically labeled images can be exported using the rendering functionalities of Blender.