Improvement of vertical height accuracy using data fusion technique for terrain mapping in oil palm plantation

• Main Author: Muhadi, Nur 'atirah
• Format: Thesis
• Language: English
• Published: 2018
• Subjects: Multisensor data fusion; Oil palm
• Online Access: http://psasir.upm.edu.my/id/eprint/75432/1/FK%202018%20117%20-%20IR.pdf; http://psasir.upm.edu.my/id/eprint/75432/
• Institution: Universiti Putra Malaysia

Digital elevation models (DEMs) play an important role in producing terrainrelated applications such as curvature and contour maps for planning and management of oil palm plantation. Compared to Light Detection and Ranging (LiDAR) data, Interferometric Synthetic Aperture Radar (IfSAR) has lower accuracy but the cost is much cheaper. In order to increase the accuracy of IfSAR data, fusion of IfSAR and terrestrial LiDAR (TLS) datasets was proposed in this study. The TLS data collection was carried out in TH Plantation in Muadzam Shah, Pahang using Faro 3D Laser Scanner. Two different stations were selected with different terrain characteristics. Station 1 was located in a relatively flat area while station 2 was located in a rolling and hilly area. Raw data of TLS were filtered using TerraScan software to extract the ground points from object points. In this study, the efficiency of filtering technique for TLS data was assessed and determined before being used for data fusion with IfSAR. The performance of data filtering was tested by using double filtering technique. Using this technique, 20,977594 points were correctly identified as object points while 10804 object points were mistakenly classified as ground points. Statistically, 0.05% of type II errors (accept object points as ground points) were obtained in the study area. The result indicates that filtering algorithm in TerraScan was good enough to be used for TLS data in oil palm plantation. When the filtering was completed, data fusion of TLS and IfSAR-derived DEM was developed to increase the accuracy of IfSAR-derived elevation models and provide high quality data for plantation management especially for slope risk management. This study used fusion by weights based on the spatial errors after applying regression equation. The results show a significant reduction in RMSEs after fusion. RMSEs of both stations reduced from 1.83 m to 0.35 m and from 3.13 m to 0.41 m for station 1 and station 2 respectively. In addition, data fusion technique for area with no TLS data that located nearby the station was tested. Data fusion of these areas was carried out by using regression equation of their relative station but the weighted values were computed differently from the previous fusion technique. The weighted value was computed using mean error of the elevation of its relative station, the mean error of the elevation based on classified elevation range and the error pattern based on its relative station. All results proved that the proposed fusion technique could be done in relatively flat area but it could not be used in steep-slope area. A mobile application was also developed for field data collection and verification. The application has been successfully developed and tested in the field. On the whole, it is concluded that data fusion is a promising technique for increasing the accuracy of IfSAR-derived DEM in oil palm plantation.