An Improved Algorithm Robust to Illumination Variations for Reconstructing Point Cloud Models from Images

Luo, Nan; Huang, Ling; Wang, Quan; Liu, Gang

doi:10.3390/rs13040567

Cited by 3 publications

(2 citation statements)

References 46 publications

(62 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We will test approaches that minimize these issues, namely the 3D reconstruction by the patch-based multiview stereo approach [50]. We will also test the most recent trend based on deep learning to deal with complex structures and smooth and less textured surfaces [51].…”

Section: Discussionmentioning

confidence: 99%

Assessment of the Influence of Survey Design and Processing Choices on the Accuracy of Tree Diameter at Breast Height (DBH) Measurements Using UAV-Based Photogrammetry

Moreira

Goyanes

Pina

et al. 2021

Drones

View full text Add to dashboard Cite

This work provides a systematic evaluation of how survey design and computer processing choices (such as the software used or the workflow/parameters chosen) influence unmanned aerial vehicle (UAV)-based photogrammetry retrieval of tree diameter at breast height (DBH), an important 3D structural parameter in forest inventory and biomass estimation. The study areas were an agricultural field located in the province of Málaga, Spain, where a small group of olive trees was chosen for the UAV surveys, and an open woodland area in the outskirts of Sofia, the capital of Bulgaria, where a 10 ha area grove, composed mainly of birch trees, was overflown. A DJI Phantom 4 Pro quadcopter UAV was used for the image acquisition. We applied structure from motion (SfM) to generate 3D point clouds of individual trees, using Agisoft and Pix4D software packages. The estimation of DBH in the point clouds was made using a RANSAC-based circle fitting tool from the TreeLS R package. All trees modeled had their DBH tape-measured on the ground for accuracy assessment. In the first study site, we executed many diversely designed flights, to identify which parameters (flying altitude, camera tilt, and processing method) gave us the most accurate DBH estimations; then, the resulting best settings configuration was used to assess the replicability of the method in the forested area in Bulgaria. The best configuration tested (flight altitudes of about 25 m above tree canopies, camera tilt 60°, forward and side overlaps of 90%, Agisoft ultrahigh processing) resulted in root mean square errors (RMSEs; %) of below 5% of the tree diameters in the first site and below 12.5% in the forested area. We demonstrate that, when carefully designed methodologies are used, SfM can measure the DBH of single trees with very good accuracy, and to our knowledge, the results presented here are the best achieved so far using (above-canopy) UAV-based photogrammetry.

show abstract

Section: Discussionmentioning

confidence: 99%

Assessment of the Influence of Survey Design and Processing Choices on the Accuracy of Tree Diameter at Breast Height (DBH) Measurements Using UAV-Based Photogrammetry

Moreira

Goyanes

Pina

et al. 2021

Drones

View full text Add to dashboard Cite

show abstract

“…In addition, the data acquisition technique to obtain data can sense from various distances (target to sensor), for instance, application of satellite imagery [19], UAV [20], and close range [21]. In addition, a dataset can be sensed under lab conditions [22], computer-generated/changed [23] or real-world (unchanged) conditions [24].…”

Section: Introductionmentioning

confidence: 99%

Two-View Structure-from-Motion with Multiple Feature Detector Operators

Dippold

Tsai

2023

Remote Sensing

View full text Add to dashboard Cite

This paper presents a novel two-view Structure-from-Motion (SfM) algorithm with the application of multiple Feature Detector Operators (FDO). The key of this study is the implementation of multiple FDOs into a two-view SfM algorithm. The two-view SfM algorithm workflow can be divided into three general steps: feature detection and matching, pose estimation and point cloud (PCL) generation. The experimental results, the quantitative analyses and a comparison with existing algorithms demonstrate that the implementation of multiple FDOs can effectively improve the performance of a two-view SfM algorithm. Firstly, in the Oxford test dataset, the RMSE reaches on average 0.11 m (UBC), 0.36 m (bikes), 0.52 m (trees) and 0.37 m (Leuven). This proves that illumination changes, blurring and JPEG compression can be handled satisfactorily. Secondly, in the EPFL dataset, the number of features lost in the processes is 21% with a total PCL of 27,673 pt, and this is only minimally higher than ORB (20.91%) with a PCL of 10,266 pt. Finally, the verification process with a real-world unmanned aerial vehicle (UAV) shows that the point cloud is denser around the edges, the corners and the target, and the process speed is much faster than existing algorithms. Overall, the framework proposed in this study has been proven a viable alternative to a classical procedure, in terms of performance, efficiency and simplicity.

show abstract

Detailed Analysis of Pedestrian Activity Recognition in Pedestrian Zones Using 3D Skeleton Joints Using LSTM

Jan,

Badella,

Berns

2024

SN COMPUT. SCI.

View full text Add to dashboard Cite

As autonomous driving technology is developing rapidly, demands for pedestrian safety, intelligence, and stability are increasing. In this situation, there is a need to discern pedestrian location and action, such as crossing or standing, in dynamic and uncertain contexts. The success of autonomous driving for pedestrian zones depends heavily on its capacity to distinguish between safe and unsafe pedestrians. The vehicles must first recognize the pedestrian, then their body movements, and understand the meaning of their actions before responding appropriately. This article presents a detailed explanation of the architecture for 3D pedestrian activity recognition using recurrent neural networks (RNN). A custom dataset was created for behaviors such as parallel and perpendicular crossing while texting or calling encountered around autonomous vehicles. A model similar to Long-Short Term Memory (LSMT) has been used for different experiments. As a result, it is revealed that the models trained independently on upper and lower body data produced better classification than the one trained on whole body skeleton data. An accuracy of 97% has been achieved for lower body and 88–90% on upper body test data, respectively.

show abstract

An Improved Algorithm Robust to Illumination Variations for Reconstructing Point Cloud Models from Images

Cited by 3 publications

References 46 publications

Assessment of the Influence of Survey Design and Processing Choices on the Accuracy of Tree Diameter at Breast Height (DBH) Measurements Using UAV-Based Photogrammetry

Assessment of the Influence of Survey Design and Processing Choices on the Accuracy of Tree Diameter at Breast Height (DBH) Measurements Using UAV-Based Photogrammetry

Two-View Structure-from-Motion with Multiple Feature Detector Operators

Detailed Analysis of Pedestrian Activity Recognition in Pedestrian Zones Using 3D Skeleton Joints Using LSTM

Contact Info

Product

Resources

About