A NEW APPROACH FOR THE CLASSIFICATION OF GROUND POINTS FROM AIRBORNE LIDAR DATA IN FORESTED AREAS NOVI PRISTUP KLASIFIKACIJI TERENSKIH TOČAKA ŠUMSKOGA PODRUČJA TEMELJEM PODATAKA PRIKUPLJENIH ZRAČNIM LASERSKIM SUSTAVOM


Kurtulgu Z., PIRTI A.

Sumarski List, cilt.149, sa.1-2, ss.45-55, 2025 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 149 Sayı: 1-2
  • Basım Tarihi: 2025
  • Doi Numarası: 10.31298/sl.149.1-2.4
  • Dergi Adı: Sumarski List
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, CAB Abstracts, Compendex, Veterinary Science Database
  • Sayfa Sayıları: ss.45-55
  • Anahtar Kelimeler: DBSCAN clustering algorithm, grid-based filter, second/last return
  • Yıldız Teknik Üniversitesi Adresli: Evet

Özet

LiDAR systems are being used increasingly widely and effectively in the field of forestry. They play an important role in many applications such as creating detailed maps of forests, obtaining high-precision elevation data, obtaining information about the height, density and distribution of trees, mapping the topography under the forest in detail, landslide and erosion control, road planning and water management. Underforest topography maps can be created with high precision thanks to LiDAR ground points. In our study, we present a new approach by using LiDAR data to create the physical characteristics of forest land. In this approach, firstly, airborne LiDAR beams were divided into datasets according returns. Three datasets were created: second return, last return and first/last return. Secondly, each dataset was positionally placed in the grid structure. Filtering was done according to mean height values of points in the cell. Thirdly, DBSCAN clustering algorithm, one of the machine learning methods, was used. The epsilon value, one of the parameters used in the DBSCAN algorithm, was determined according to the silhouette index, and LiDAR ground points were classified. The classified LiDAR ground points were compared with the existing ground control points. As a result, the combination of the second return and the last return dataset showed successful results with a kappa value of 82.27% and an F1 score value of 0.71. Also, digital terrain models were created and compared. To demonstrate the effectiveness of the proposed approach, data was compared with the CSF algorithm, which is one of the traditional filtering methods. After the accuracy evaluations, we were able to classify more LiDAR ground points with our proposed approach. Thus, we think that LiDAR ground data can create a detailed and accurate topography map, define forest features and contribute to the decision-making process for forestry activities.