Staff

Abstract:
This paper describes a software toolbox for LiDAR (Light Detection and Ranging) and LiDAR-Inertial Odometry qualitative and quantitative evaluation. We provide software as https://github.com/MapsHD organization with all necessary information at https://github.com/MapsHD/HDMapping. Our software contributions are a) ground truth data processing tool, b) dockerized state-of-the-art LO and LIO algorithms, c) multi-session data registration to common coordinate system, d) Absolute Pose Error (APE) and Relative Pose Error (RPE) metrics, e) import/export tools for easier 3D data handling and visualizing, e.g., in Cloud Compare software. This software is compatible with ROS1 (Robot Operating System) and ROS2 data formats. We show an example benchmark of LeGO-LOAM, LIO-SAM, FAST-LIO, DLO, VoxelMap, Faster-LIO, KISS-ICP, CT-ICP, SLICT, DLIO, GLIM, iG-LIO, LIO-EKF, I2EKF-LO, GenZ-ICP, RESPLE, odometry_ros_wrapper, Point-LIO, and LOAM-Livox algorithms. For all experiments we provide movies. The contribution of the paper is software-oriented LO/LIO algorithm benchmark suite. The novelty lies in the integration of multiple benchmarking steps into a unified framework, thus overall effort needed for qualitative and quantitative evaluation is reduced.

Keywords:
LiDAR odometry, LiDAR-inertial odometry, Benchmarking

Abstract:
[Purpose of Review:] This paper provides an overview of integrating artificial intelligence (AI), particularly deep learning (DL), with ground-based LiDAR point clouds for forest monitoring. It identifies trends, highlights advancements, and discusses future directions for AI-supported forest monitoring.

[Recent Findings:] Recent studies indicate that DL models significantly outperform traditional machine learning methods in forest inventory tasks using terrestrial LiDAR data. Key advancements have been made in areas such as semantic segmentation, which involves labeling points corresponding to different vegetation structures (e.g., leaves, branches, stems), individual tree segmentation, and species classification. Main challenges include a lack of standardized evaluation metrics, limited code and data sharing, and reproducibility issues. A critical issue is the need for extensive reference data, which hinders the development and evaluation of robust AI models. Solutions such as the creation of large-scale benchmark datasets and the use of synthetic data generation are proposed to address these challenges. Promising AI paradigms like Graph Neural Networks, semi-supervised learning, self-supervised learning, and generative modeling have shown potential but are not yet fully explored in forestry applications.

[Summary:] The review underscores the transformative role of AI, particularly DL, in enhancing the accuracy and efficiency of forest monitoring using ground-based 3D point clouds. To advance the field, there is a critical need for comprehensive benchmark datasets, open-access policies for data and code, and the exploration of novel DL architectures and learning paradigms. These steps are essential for improving research reproducibility, facilitating comparative studies, and unlocking new insights into forest management and conservation.

Keywords:
Deep learning, Machine learning, Forest inventory, Tree characteristics, Open data, Precision forestry, LiDAR, TLS

Abstract:
Mobile devices such as smartphones, tablets or similar devices are becoming increasingly important as measurement devices in forestry due to their advanced sensors, including RGB cameras and LiDAR systems. This review examines the current state of applications of mobile devices for measuring biometric characteristics of individual trees and presents technologies, applications, measurement accuracy and implementation barriers. Passive sensors, such as RGB cameras have proven their potential for 3D reconstruction and analysing point clouds that improve single treelevel information collection. Active sensors with LiDAR-equipped smartphones provide precise quantitative measurements but are limited by specific hardware requirements. The combination of passive and active sensing techniques has shown significant potential for comprehensive data collection. The methods of data collection, both physical and digital, significantly affect the accuracy and reproducibility of measurements. Applications such as ForestScanner and TRESTIMATM have automated the measurement of tree characteristics and simplified data collection. However, environmental conditions and sensor limitations pose a challenge. There are also computational obstacles, as many methods require significant post-processing. The review highlights the advances in mobile device-based forestry applications and emphasizes the need for standardized protocols and cross-device benchmarking. Future research should focus on developing robust algorithms and cost-effective solutions to improve measurement accuracy and accessibility. While mobile devices offer significant potential for forest surveying, overcoming the above-mentioned challenges is critical to optimizing their application in forest management and protection.

Keywords:
mobile device, tree attributes , LiDAR, Photogrammetry

Abstract:
Precise forest inventory in difficult terrain remains challenging due to mobility constraints and canopy occlusion. This paper presents an affordable air-ground mobile mapping system combining shoulder-mounted double LiDAR with a lightweight FPV drone-based LiDAR. We developed an extrinsic calibration method for dual orthogonally-mounted sensors and implemented a comprehensive processing pipeline incorporating LiDAR odometry, pose graph SLAM, and multi-view Normal Distributions Transform. Field ex-periments demonstrate successful air-ground data fusion for tree stem detection and dendrometric parameter extraction. The system was validated in extreme environments including a cave survey, proving versatility in difficult terrain. All software components are released as open-source tools: https://github.com/MapsHD/HDMapping.

Keywords:
UAV LiDAR, Forest Inventory, Under-Canopy Mapping, SLAM, Open Source

Keywords:
leśnictwo, sztuczna inteligencja, uczenie maszynowe