UAV-based extraction of topograhic and as-built information by object-based image analysis technique
The advancement of airborne technology without pilot, unmanned aerial vehicle (UAV) systems utilises the minimal cost of function for future mapping purposes. The utilisation of UAV data from visible red, green, blue (RGB) bands is limited to the visualisation of orthophoto for planning and monitori...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/76072/1/FK%202018%20151%20-%20IR.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The advancement of airborne technology without pilot, unmanned aerial vehicle (UAV) systems utilises the minimal cost of function for future mapping purposes. The utilisation of UAV data from visible red, green, blue (RGB) bands is limited to the visualisation of orthophoto for planning and monitoring applications. Thus, this study explores the potentials of UAV data based on RGB camera for topographic and as-built information for features extraction using OBIA (objectbased image analysis) technique. The main objective of this study is to assess the capability of UAV data in providing reliable topographic and as-built data information. More specifically, this study aims to extract topographic and as-built information such as land cover and geometry of infrastructure classes in urban area using eBee Sensefly UAV imagery. In this frame of study, The National Land and Survey Institute in Tanjung Malim, Perak, Malaysia was chosen as the area of interest. A robust Taguchi method was used in optimising the segmentation process. In accordance with the image classification process, different supervised OBIA classifiers such as KNN, normal Bayes (NB), decision tree (DT), random forest (RF), and support vector machine (SVM) were tested by tuning each of their parameters to quantify the performance of each classifier in favour of using UAV image data. Results showed that SVM obtained the highest percentage of overall accuracy, followed by RF, NB, DT, and KNN at 97.20%, 95.80%, 93.14%, 86.01% and 77.62%, respectively. The optimal OBIA parameters for each classifier are as follows: SVM with their C was 100,000 and Gamma 0.001. Meanwhile, the maximum depth parameter for RF and DT was 15 and 20 respectively. KNN classifier with K parameter was 5. The dimensional assessment on features that was extracted using OBIA showed that the error was in the range of less than 20 cm. Meanwhile, the positional accuracy based on root mean square error (RMSE) result gave the error of horizontal and vertical axes of less than 0.5 m. This result indicated that the UAV image data have a big potential to be utilised for topographic mapping and as-built information. Moreover, the OBIA technique also contributes to efficient features extraction compared to the manually practiced technique. |
|---|