LidarCSNet: A Deep Convolutional Compressive Sensing Reconstruction Framework for 3D Airborne Lidar Point Cloud
Graphical abstract
Section snippets
Introduction and related work
Lidar (Light Detection and Ranging) is an optical remote sensing data acquisition technique that produces 3D data with highly accurate measurements. The x , y, and z measurements are derived from the Time of Travel (ToT) after applying mathematical transformations. Lidar scanner works similarly to the principle of Electronic Distance Measuring Instrument (EDMI). In EDMI, a laser pulse is fired from the transmitter of the transceiver and the reflected energy of the pulse is captured by the
Methods and materials
We describe dataset details and acquired methodology for our proposed LidarCSNet, LidarNet, and LidarNet++ in the following sections.
Experiments and results
We present comprehensive results obtained for extensive experiments based on the two case studies for forest and urban environment in the coming sections.
Conclusion and future directions
Lidar point cloud data has been widely used in research as well as industrial applications. Geospatial applications involving analysis of forest dynamics, urban environment mapping and surveying uses lidar point cloud data. In our work, we are proposing a deep Convolutional Neural Network based Compressive Sensing framework, named LidarCSNet, to generate sparse approximation of the lidar point cloud data for selected {4%, 10%, 25%, 50% and 75%} CS measurement ratios. We perform multiple
CRediT authorship contribution statement
Rajat C. Shinde: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing - original draft, Visualization. Surya S. Durbha: Conceptualization, Methodology, Resources, Data curation, Writing - review & editing, Supervision, Project administration. Abhishek V. Potnis: Methodology, Resources.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The authors express their gratitude towards the Training Center for Applied Geodesy and Photogrammetry (UP TCAGP) and the PHIL-Lidar Program of Philippines for publishing the open lidar data. The authors also express their gratitude to the Google Cloud Team for providing the Google Cloud Platform with GPU enabled computing facility for implementing the architectures under the Google Cloud Platform Research Credits Program.
Funding
This research is partially funded under the Prime Minister’s Research Fellowship issued by the Ministry of Education, Government of India. The high-performance computation for lidar data processing is performed using the Google Cloud Platform Credits received under the Google Cloud Platform Research Credits Program.
References (77)
- et al.
A study of forest biomass estimates from lidar in the northern temperate forests of New England
Remote Sens. Environ.
(2013) - et al.
The importance of spatial detail: Assessing the utility of individual crown information and scaling approaches for lidar-based biomass density estimation
Remote Sens. Environ.
(2015) - et al.
Above ground biomass estimation in an African tropical forest with lidar and hyperspectral data
ISPRS J. Photogramm. Remote Sens.
(2014) - et al.
A survey of mobile laser scanning applications and key techniques over urban areas
Remote Sens.
(2019) - Abadi, M., Barham, P., Chen, J., Chen, Z., Davis, A., Dean, J., Devin, M., Ghemawat, S., Irving, G., Isard, M., Kudlur,...
- Adler, A., Boublil, D., Zibulevsky, M., 2017. Block-Based compressed sensing of images via deep learning, in: 2017 IEEE...
- Alonzo, M., Bookhagen, B., McFadden, J.P., Sun, A., Roberts, D.A., 2015. Mapping urban forest leaf area index with...
- et al.
Compressive sensing based three-dimensional imaging method with electro-optic modulation for nonscanning laser radar
Symmetry (Basel).
(2020) - Andaya, K.J., Alviar, J., Mars, P., Cruz, C., Sarmiento, C.J., Balicanta, L., Paringit, E., 2015. Airborne LiDAR...
- Andersen, H.E., McGaughey, R.J., Reutebuch, S.E., 2005. Estimating forest canopy fuel parameters using LIDAR data....
Isolating individual trees in a savanna woodland using small footprint lidar data
Photogramm. Eng. Remote Sensing
A fast reconstruction method of the dense point-cloud model for cultural heritage artifacts based on compressed sensing and sparse auto-encoder
Opt. Quantum Electron.
Digital surface models and building extraction: a comparison of IFSAR and LIDAR data
IEEE Trans. Geosci. Remote Sens. doi
The LAS 1.4 Specification
Photogramm. Eng. Remote Sensing.
Deep learning-based tree classification using mobile LiDAR data
Remote Sens. Lett.
Deep learning for 3D point clouds: A survey
IEEE Trans. Pattern Anal. Mach. Intell.
A segmentation-based method to retrieve stem volume estimates from 3-D tree height models produced by laser scanners
IEEE Trans. Geosci. Remote Sens. doi
Deep fully-connected networks for video compressive sensing
Digit. Signal Process. A Rev. J.
Comparison of surface normal estimation methods for range sensing applications
IEEE International Conference on Robotics and Automation
ReconNet: Non-iterative reconstruction of images from compressively sensed measurements
Theory and applications of compressed sensing
GAMM Mitteilungen
Cited by (15)
GCN-based proximal unrolling matrix completion for piecewise smooth signal recovery
2023, Signal ProcessingUltrasmall fully-convolution GVA-net for point cloud processing[Formula presented]
2023, Applied Soft ComputingCitation Excerpt :As the availability and accuracy of 3D scanners, LiDAR tools, and photogrammetry software s, the use of point clouds is becoming a dominant approach in a variety of industrial branches. Deformation detection in civil engineering [1], 3D roof reconstruction [2], forest dynamic observations [3], or disaster damage detection [4] are just cases in point. Among tasks related to point clouds, we may differentiate two groups: semantic tasks and auxiliary tasks.
Fore-Net: Efficient inlier estimation network for large-scale indoor scenario
2022, ISPRS Journal of Photogrammetry and Remote SensingCitation Excerpt :With the increasing maturity of three-dimensional (3D) scanning technology, 3D sensors such as Lidar or depth cameras have become the primary option for sensing the environment. Point cloud registration, which is the process of finding rigid transformation between different scans of a target scene, has become a critical problem in 3D reconstruction (Shinde et al., 2021; Jung et al., 2014), autonomous driving (Li and Ibanez-Guzman, 2020; Li et al., 2021), robot positioning (Röwekämper et al., 2012; Lee et al., 2019), and many other applications. However, when used in the real-world environment, especially in cluttered indoor spaces, point clouds scanned by sensors are often corrupted by noise because of sensor limitations, occlusions, or artifacts.
Monitoring of Levee Deformation for Urban Flood Risk Management Using Airborne 3D Point Clouds
2024, Water (Switzerland)