Predicting Maps Using In-Vehicle Cameras for Data-Driven Intelligent Transport DOI Open Access
Zhiguo Ma, Y Zhang, Meng Han

et al.

Electronics, Journal Year: 2023, Volume and Issue: 12(24), P. 5017 - 5017

Published: Dec. 15, 2023

Bird’s eye view (BEV) semantic maps have evolved into a crucial element of urban intelligent traffic management and monitoring, offering invaluable visual significant data representations for informed city decision making. Nevertheless, current methodologies continue underutilizing the temporal information embedded within dynamic frames throughout BEV feature transformation process. This limitation results in decreased accuracy when mapping high-speed moving objects, particularly capturing their shape trajectory. A framework is proposed cross-view segmentation to address this challenge, leveraging simulated environments as starting point before applying it real-life imaginative transportation scenarios. The converter module thoughtfully designed collate from multiple initial observations captured various angles modes. outputs top-down graph characterized by its object space layout preserve beneficial transformation. NuScenes dataset used evaluate model effectiveness. novel application also devised that harnesses transformer networks map images video sequences or comprehensive bird’s-eye views. By combining physics-based constraint-based formulations conducting ablation studies, approach has been substantiated, highlighting significance context above below given generating these maps. innovative method thoroughly validated on dataset. Notably, yielded state-of-the-art instantaneous results, with particular benefits observed smaller category displays. experimental findings include comparing axial attention (SOTA) model, demonstrating performance enhancement associated awareness.

Language: Английский

A Novel Approach for Simultaneous Localization and Dense Mapping Based on Binocular Vision in Forest Ecological Environment DOI Open Access
Lina Liu, Yaqiu Liu, Yunlei Lv

et al.

Forests, Journal Year: 2024, Volume and Issue: 15(1), P. 147 - 147

Published: Jan. 10, 2024

The three-dimensional reconstruction of forest ecological environment by low-altitude remote sensing photography from Unmanned Aerial Vehicles (UAVs) provides a powerful basis for the fine surveying resources and management. A stereo vision system, D-SLAM, is proposed to realize simultaneous localization dense mapping UAVs in complex environments. system takes binocular images as input 3D maps target outputs, while sparse camera poses can be obtained. tracking thread utilizes temporal clue match map points zero-drift localization. relative motion amount data association between frames are used constraints new keyframes selection, image spatial compensation strategy increase robustness algorithm tracking. uses Linear Attention Network (LANet) predict reliable disparity ill-posed regions, which transformed depth constructing point cloud maps. Evaluations three datasets, EuRoC, KITTI Forest, show that run at 30 ordinary 3 per second with high accuracy several centimeters Root Mean Squared Absolute Trajectory Error (RMS ATE) on EuRoC Relative (RMSE) two average values 0.64 0.2 trel Rrel KITTI, outperforming most mainstream models terms robustness. Moreover, advantage compensates shortcomings Smultaneous Localization Mapping (SLAM) systems meets requirements real-time forests.

Language: Английский

Citations

5
Evaluation of Handheld Mobile Laser Scanner Systems for the Definition of Fuel Types in Structurally Complex Mediterranean Forest Stands DOI Creative Commons
Raúl Hoffrén, María Teresa Lamelas Gracia, Juan de la Riva

et al.

Fire, Journal Year: 2024, Volume and Issue: 7(2), P. 59 - 59

Published: Feb. 18, 2024

The exposure of Mediterranean forests to large wildfires requires mechanisms prevent and mitigate their negative effects on the territory ecosystems. Fuel models synthesize complexity heterogeneity forest fuels allow for understanding modeling fire behavior. However, it is sometimes challenging define fuel type in a structurally heterogeneous stand due mixture characteristics from different types limitations qualitative field observations passive active airborne remote sensing. This can impact performance classification that rely situ identification as ground truth, which lead mistaken prediction over larger areas models. In this study, handheld mobile laser scanner (HMLS) system was used assess its capability Prometheus 43 plots Aragón (NE Spain). HMLS captured vertical horizontal distribution at an extremely high resolution derive high-density three-dimensional point clouds (average: 63,148 points/m2), were discretized into voxels 0.05 m3. total number each 5 cm height stratum calculated quantify volume stratum, providing (m3/m2) plot centimetric scale. Additionally, computed (0.60, 2, 4 m) plot. satisfactorily identified compared with estimated field. led modification truth 10 out plots, resulting errors being found estimation between FT2–FT3, FT5–FT6, FT6–FT7. These results demonstrate ability systems capture scales definition forests, making them powerful tools mapping, modeling, ultimately improving wildfire prevention management.

Language: Английский

Citations

4
Depth Estimation from a Hierarchical Baseline Stereo with a Developed Light Field Camera DOI Creative Commons
Fei Liu,

Guangqi Hou

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(2), P. 550 - 550

Published: Jan. 8, 2024

This paper presents a hierarchical baseline stereo-matching framework for depth estimation using novelly developed light field camera. The imaging process of micro-lens array-based camera is derived. A macro-pixel map constructed by treating each as one in the field’s raw image. For macro-pixel, feature vector represented leveraging texture and gradient cues over surrounding ring neighboring macro-pixels. Next, micro-lenses containing edges are detected on map. Hierarchical performed macro-pixel-wise coarse matching pixel-wise fine matching, effectively eliminating ambiguities. Finally, post-processing step applied to improve accuracy. lab-designed camera’s performance evaluated terms accuracy processing speed capturing real-world scenes under studio lighting conditions. And an experiment rendered synthetic samples conducted quantitative evaluation, showing that maps with local details can be accurately recovered.

Language: Английский

Citations

3
Noise Analysis for Unbiased Tree Diameter Estimation from Personal Laser Scanning Data DOI Creative Commons
Karel Kuželka, Peter Surový

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(7), P. 1261 - 1261

Published: April 2, 2024

Personal laser scanning devices employing Simultaneous Localization and Mapping (SLAM) technology have rightfully gained traction in various applications, including forest mensuration inventories. This study focuses the inherent stochastic noise SLAM data. An analysis of distribution is performed GeoSLAM ZEB Horizon for point clouds trees two species, Norway spruce European beech, to mitigate bias diameter estimates. The method involved evaluating residuals individual 3D points concerning real tree surface model based on TLS results show that not symmetrical regarding surface, showing significant negative difference, moreover, difference from zero mean significantly differs between with an average −0.40 cm −0.44 beech. Furthermore, dependence return distance scanner target incidence angle. experimental comparison RANSAC circle fitting outcomes under configurations showed unbiased estimates extending inlier tolerance 5 2.5 asymmetry. By nonvalidity assumption estimation methods, contribute fill a gap methodology data processing widely utilized instrument.

Language: Английский

Citations

2
3D LiDAR Multi-Object Tracking with Short-Term and Long-Term Multi-Level Associations DOI Creative Commons
Minho Cho, Euntai Kim

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(23), P. 5486 - 5486

Published: Nov. 24, 2023

LiDAR-based Multi-Object Tracking (MOT) is a critical technology employed in various autonomous systems, including self-driving vehicles and delivery robots. In this paper, novel 3D MOT approach introduced. The proposed method was built upon the Tracking-by-Detection (TbD) paradigm incorporated multi-level associations that exploit an object’s short-term long-term relationships with existing tracks. Specifically, association leverages fact objects do not move much between consecutive frames. contrast, assesses degree to which trajectory aligns current detections. evaluation of matching detection maintained performed using Graph Convolutional Network (GCN). Furthermore, inactive track address issue incorrect ID switching for have been occluded extended period. evaluated on KITTI benchmark tracking dataset achieved Higher-Order Accuracy (HOTA) 75.65%, marking 5.66% improvement over AB3DMOT, while also accomplishing number switches 39, 74 less than AB3DMOT. These results confirmed effectiveness diverse road environments.

Language: Английский

Citations

4
Individual Tree Identification and Segmentation in Pinus spp. Stands through Portable LiDAR DOI Open Access
Ana Solares-Canal, Laura Alonso, Juan Picos

et al.

Forests, Journal Year: 2024, Volume and Issue: 15(7), P. 1133 - 1133

Published: June 28, 2024

Forest inventories are essential for sustainable forest management. In at the tree level, all information is linked to individuals: species, diameter, height, or spatial distribution, example. Currently, implementation of Portable LiDAR (PLS) being studied, aiming digitalize environments and increase reliability observations. Performing automatic individual identification (ITD) segmentation (ITS) operational PLS in forestry. Multiple algorithms have been developed performing these tasks point clouds. Their performance varies according system characteristics stand. this study, several ITD ITS analyzed very high-density clouds Pinus species stands with a varying presence understory, shrubs, branches. The results showed that methods based on finding trunks more suitable regular no understory. process, evaluated highly conditioned by understory comparison help identify most algorithm be applied types stands, hence, they might enhance operability systems.

Language: Английский

Citations

1
Evaluating the efficacy of sampling acquisition paths for mapping vegetation structure using terrestrial mobile laser scanning. DOI Creative Commons

Johann Tiede,

Karin Reinke, Simon Jones

et al.

Ecological Informatics, Journal Year: 2024, Volume and Issue: 82, P. 102675 - 102675

Published: June 15, 2024

Mobile Laser Scanning (MLS) is rapidly increasing in popularity for the capture and quantification of vegetation structure natural landscapes. However, understanding optimal implementation forest field still limited, particularly regarding choice acquisition path length followed during data capture. How variable characterisation when using different paths, this difference likely to be significant most users? In study we compared four designs commonly cited literature determine importance users capturing structure. MLS point clouds were systematically captured repeatedly survey plots a closed-canopy ecosystem south-eastern Australia. Digital elevation models, canopy height vertical voxel occupancy profiles derived illustrate sensitivity path's configuration variability We found strong agreement between that increased as increased. No differences digital models at 75% 100% lengths. observed two 25% 50% respectively. Significant model only one plot length. Mean reference was 0.09 m across all plots, 0.4 models. Voxel showed greatest understory where mean paths 5.72%. Our findings have implications use from an operational perspective, they reliability their forested Users can choose balance design with meet requirements such efficiency or ease field.

Language: Английский

Citations

1
Automatic Tree Detection/Localization in Urban Forest Using Terrestrial Lidar Data DOI
Renato César dos Santos, Matheus Ferreira Da Silva, Antônio Maria Garcia Tommaselli

et al.

IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium, Journal Year: 2024, Volume and Issue: XXXVI, P. 4522 - 4525

Published: July 7, 2024

Language: Английский

Citations

1
Remote Sensing Parameter Extraction of Artificial Young Forests under the Interference of Undergrowth DOI Creative Commons

Zefu Tao,

Lubei Yi,

Z. F. Wang

et al.

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(17), P. 4290 - 4290

Published: Aug. 31, 2023

Due to the lower canopy height at maximum crown width bottom of young Picea crassifolia trees, they are mixed with undergrowth. This makes it challenging accurately determine size using CHM data or point cloud data. UAV imagery, on other hand, incorporates rich color information and, after processing, can effectively distinguish between spruce trees and ground vegetation. In this study, experimental site was an artificial forest in Shangshan Village, Qinghai Province, China. images were used obtain normalized saturation for sample plots. A marker-controlled watershed segmentation algorithm employed extract tree parameters, results compared those obtained via clustering based Canopy Height Model (CHM) images. The research showed that single recognition capabilities three types similar, F-measures 0.96, 0.95, 0.987 image, data, respectively. mean square errors extracted from imagery 0.043, 0.125, 0.046 plots, which better than values 0.103, 0.182, 0.074 as well 0.36, 0.461, 0.4 exhibited fitting extraction result indicates UAV-acquired optical has applicability extracting individual feature parameters compensate deficiencies

Language: Английский

Citations

1
Evaluation of Handheld Mobile Laser Scanner Systems for the Definition of Fuel Types in Structurally Complex Mediterranean Forest Stands DOI Open Access
Raúl Hoffrén, María Teresa Lamelas Gracia, Juan de la Riva

et al.

Published: Dec. 26, 2023

The exposure of Mediterranean forests to large wildfires requires mechanisms prevent and mitigate their negative effects on the territory ecosystems. Fuel models synthesize complexity heterogeneity forest fuels allow understanding modelling fire behavior. However, it is sometimes challenging define fuel type in a structurally heterogeneous stand due mixture characteristics from different types limitations qualitative field observation passive active airborne remote sensing. This can impact performance classification that rely situ identification as ground-truth, which lead mistaken prediction over larger areas models. In this study, handheld mobile laser scanner (HMLS) system were used assess its capability Prometheus 43 plots Aragón (NE Spain). HMLS captured vertical horizontal distribution at extremely high resolution derive high-density three-dimensional point clouds (average: 63,148 points/m2) discretized into voxels 0.05 m3. total number each 5 cm height stratum was calculated quantify volume stratum, providing (m3/m2) for plot centimetric scale. Additionally, computed (0.60, 2, 4 m) plot. satisfactory identified plot, compared with estimated field. led modification ground-truth 10 out plots, estimation errors between 2–3, 5–6, 6–7. These results demonstrate ability systems capture scales definition forests, making them powerful tools mapping modelling, ultimately improving wildfire prevention management.

Language: Английский

Citations

1