Heidar Rastiveis

Backpack LiDAR systems are gaining popularity due to their rapid data acquisition, portability, and cost-effectiveness. However, using backpack LiDAR in forest poses challenges, such as GNSS signal attenuation under the canopy, leading to inaccurate trajectory estimates and misregistered point clouds. This paper aims at presenting a novel method that addresses this challenge by leveraging the integrated scans (IS) concept to enhance point cloud quality for automated forest inventory. The proposed method for forest inventory using ISs consists of four key steps: (i) Integrated scans (IS) generation, (ii) feature extraction, matching, and tracking, (iii) trajectory enhancement, and (iv) tree biometric extraction. Firstly, IS point clouds are generated based on the initial GNSS/INS trajectory. Secondly, reliable forest features such as tree trunks and ground patches are extracted, matched, and tracked across ISs. These features are then utilized in the trajectory enhancement step, where a non-linear Least Squares Adjustment (LSA) technique is used to minimize discrepancies among the features to enhance the trajectory. The resulting point clouds, based on the improved trajectory, are used to extract tree biometric information. The proposed method was evaluated using two distinct datasets collected with different systems. The evaluation results, both qualitatively and quantitatively, validate the effectiveness of the proposed method, showcasing its potential for fine-scale forest inventory applications.

Mobile terrestrial laser scanning (MTLS) systems provide a safe and efficient means to survey roadway corridors at high speed. MTLS point clouds are rich in planimetric data. However, manual extraction of useful information from these point clouds can be time consuming and laborious and automated object extraction from MTLS point clouds has become a hot topic in the remote sensing community. This study proposes an automated method for power line extraction from MTLS point clouds based on a multilayer perceptron (MLP) neural network. The proposed method consists of three main steps: (i) point cloud preprocessing, (ii) descriptor extraction and selection, and (iii) point classification. The preprocessing step involves filtering out more than 90% of the point cloud by eliminating the vast majority of unneeded points. Next, various descriptors are extracted from the remaining points including planarity, linearity, and verticality, and the descriptor standard deviation is used to select the best‐suited descriptors for power line extraction. Finally, an MLP neural network is trained using the selected descriptors from several cable and noncable sample points. The proposed algorithm was evaluated in three MTLS point clouds in urban and nonurban environments totalling 5.5 kilometres in length. An average precision of 94% and a recall of 94% showed the algorithm’s reliability and feasibility.

The 3D semantic segmentation of a LiDAR point cloud is essential for various complex infrastructure analyses such as roadway monitoring, digital twin, or even smart city development. Different geometric and radiometric descriptors or diverse combinations of point descriptors can extract objects from LiDAR data through classification. However, the irregular structure of the point cloud is a typical descriptor learning problem—how to consider each point and its surroundings in an appropriate structure for descriptor extraction? In recent years, convolutional neural networks (CNNs) have received much attention for automatic segmentation and classification. Previous studies demonstrated deep learning models’ high potential and robust performance for classifying complicated point clouds and permutation invariance. Nevertheless, such algorithms still extract descriptors from independent points without investigating the deep descriptor relationship between the center point and its neighbors. This paper proposes a robust and efficient CNN-based framework named D-Net for automatically classifying a mobile laser scanning (MLS) point cloud in urban areas. Initially, the point cloud is converted into a regular voxelized structure during a preprocessing step. This helps to overcome the challenge of irregularity and inhomogeneity. A density value is assigned to each voxel that describes the point distribution within the voxel’s location. Then, by training the designed CNN classifier, each point will receive the label of its corresponding voxel. The performance of the proposed D-Net method was tested using a point cloud dataset in an urban area. Our results demonstrated a relatively high level of performance with an overall accuracy (OA) of about 98% and precision, recall, and F1 scores of over 92%.

The utilization of remote sensing technologies for archaeology was motivated by their ability to map large areas within a short time at a reasonable cost. With recent advances in platform and sensing technologies, uncrewed aerial vehicles (UAV) equipped with imaging and Light Detection and Ranging (LiDAR) systems have emerged as a promising tool due to their low cost, ease of deployment/operation, and ability to provide high-resolution geospatial data. In some cases, archaeological sites might be covered with vegetation, which makes the identification of below-canopy structures quite challenging. The ability of LiDAR energy to travel through gaps within vegetation allows for the derivation of returns from hidden structures below the canopy. This study deals with the development and deployment of a UAV system equipped with imaging and LiDAR sensing technologies assisted by an integrated Global Navigation Satellite System/Inertial Navigation System (GNSS/INS) for the archaeological mapping of Dana Island, Turkey. Data processing strategies are also introduced for the detection and visualization of underground structures. More specifically, a strategy has been developed for the robust identification of ground/terrain surface in a site characterized by steep slopes and dense vegetation, as well as the presence of numerous underground structures. The derived terrain surface is then used for the automated detection/localization of underground structures, which are then visualized through a web portal. The proposed strategy has shown a promising detection ability with an F1-score of approximately 92%.

Adequate water pavement surface drainage on highways is crucial in minimizing the potential of hydroplaning. Highway cross slope has a significant effect of draining water laterally from the pavement surface. Currently, field surveying techniques and other manual measurement methods are used to collect cross slope data on a limited basis in most states, despite these methods being labor intensive and exposing personnel to traffic. Furthermore, field surveying techniques cannot provide continuous data and can only be conducted at sample-based locations. This study conducted a technical evaluation of the effectiveness of airborne LiDAR (light detection and ranging) scanning and mobile terrestrial LiDAR scanning systems in measuring pavement cross slopes. Cross slope data were extracted from the LiDAR point cloud at five selected test sections using two different methods: (i) end-to-end method using elevations only from the pavement edge lines to generate the cross slope; and (ii) 0.2 ft interval point extraction along the cross-section and using a fitted linear regression line as the basis for the cross slope. Cross slopes were also measured at test section locations using conventional surveying methods and compared with LiDAR-extracted cross slopes. Results demonstrate that LiDAR methods are reliable for collecting accurate pavement cross slopes and should be considered for the purpose of cross slope verification on a braod scale such as statewide to address cross slope and pavement surface drainage issues proactively.

Building damage maps can be generated from either optical or Light Detection and Ranging (Lidar) datasets. In the wake of a disaster such as an earthquake, a timely and detailed map is a critical reference for disaster teams in order to plan and perform rescue and evacuation missions. Recent studies have shown that, instead of being used individually, optical and Lidar data can potentially be fused to obtain greater detail. In this study, we explore this fusion potential, which incorporates deep learning. The overall framework involves a novel End-to-End convolutional neural network (CNN) that performs building damage detection. Specifically, our building damage detection network (BDD-Net) utilizes three deep feature streams (through a multi-scale residual depth-wise convolution block) that are fused at different levels of the network. This is unlike other fusion networks that only perform fusion at the first and the last levels. The performance of BDD-Net is evaluated under three different phases, using optical and Lidar datasets for the 2010 Haiti Earthquake. The three main phases are: (1) data preprocessing and building footprint extraction based on building vector maps, (2) sample data preparation and data augmentation, and (3) model optimization and building damage map generation. The results of building damage detection in two scenarios show that fusing the optical and Lidar datasets significantly improves building damage map generation, with an overall accuracy (OA) greater than 88%.

Traditional mapping and monitoring of agricultural fields are expensive, laborious, and may contain human errors. Technological advances in platforms and sensors, followed by artificial intelligence (AI) and deep learning (DL) breakthroughs in intelligent data processing, led to improving the remote sensing applications for precision agriculture (PA). Therefore, technological advances in platforms and sensors and intelligent data processing methods, such as machine learning and DL, and geospatial and remote sensing technologies, have improved the quality of agricultural land monitoring for PA needs. However, providing ground truth data for model training is a time-consuming and tedious task and may contain multiple human errors. This paper proposes an automated and fully unsupervised framework based on image processing and DL methods for plant detection in agricultural lands from very high-resolution drone remote sensing imagery. The proposed framework’s main idea is to automatically generate an unlimited amount of simulated training data from the input image. This capability is advantageous for DL methods and can solve their biggest drawback, i.e., requiring a considerable amount of training data. This framework’s core is based on the faster regional convolutional neural network (R-CNN) with the backbone of ResNet-101 for object detection. The proposed framework’s efficiency was evaluated by two different image sets from two cornfields, acquired by an RGB camera mounted on a drone. The results show that the proposed method leads to an average counting accuracy of 90.9%. Furthermore, based on the average Hausdorff distance (AHD), an average object detection localization error of 11 pixels was obtained. Additionally, by evaluating the object detection metrics, the resulting mean precision, recall, and F1 for plant detection were 0.868, 0.849, and 0.855, respectively, which seem to be promising for an unsupervised plant detection method.