- Titel
- Development of a Material Classification Model for Multispectral LiDAR Data
- Zitierfähige Url:
- https://nbn-resolving.org/urn:nbn:de:bsz:ch1-qucosa2-1033915
- Datum der Einreichung
- 17.06.2024
- DOI
- https://doi.org/10.60687/2026-0080
- Abstract (EN)
- The advance of the new generation of LiDAR systems, with the capability of measuring the spectrum, suffers from the lack of available hardware and the elevated costs of supercontinuum white light laser systems. This work addresses this issue by evaluating the feasibility of creating a spectral dataset with target materials using a commercially available FT-IR spectrometer and applying machine learning with embedded feature selection techniques to identify the most informative wavelengths that allow the classification of materials using lower spectral resolution. Here, the performance of versions of L1-regularized logistic regression and random forest with recursive feature elimination are compared for up to 100 wavelengths. The results show a superiority of the random forest in accuracy for all the versions, with a selected version achieving an accuracy 30.4% higher than the L1-regularized logistic regression while using one less feature. This advantage comes with the drawback of predicting the material of one spectrum 60 times slower. In the end, the challenges of the multispectral LiDAR hardware are discussed, and the random forest version with 5 wavelengths is tested in an example scenario created in a laboratory multispectral LiDAR demonstrator. The results show that this approach is promising for reducing hardware costs through the use of discrete laser systems, as the selected model achieved an accuracy of 84.77% in the scenario created using the amplitude of only 5 wavelengths.
- Freie Schlagwörter (EN)
- LiDAR, multispectral, material, classification
- Klassifikation (DDC)
- 000
- Normschlagwörter (GND)
- Lidar, Material, Klassifikation, Hardware, Kosten
- BetreuerIn Hochschule / Universität
- Prof. Dr. Wolfram Hardt
- Dr. Ariane Heller
- BetreuerIn - externe Einrichtung
- Dipl.-Ing. Marco Meinig
- Den akademischen Grad verleihende / prüfende Institution
- Technische Universität Chemnitz, Chemnitz
- Version / Begutachtungsstatus
- angenommene Version / Postprint / Autorenversion
- URN Qucosa
- urn:nbn:de:bsz:ch1-qucosa2-1033915
- Veröffentlichungsdatum Qucosa
- 25.03.2026
- Dokumenttyp
- Masterarbeit / Staatsexamensarbeit
- Sprache des Dokumentes
- Englisch
- Lizenz / Rechtehinweis
