IMPROVEMENT OF SPATIAL DATA QUALITY IN PBB USING COORDINATES TRANSFORMATION BASED ON ITERATIVE CLOSEST POINT (ICP) ALGORITHM TO OBJECT SEGMENTATION ON QUICKBIRD IMAGE
ABSTRACT: <br /> <br /> <br /> <br /> <br /> Due to different acquisition method and not synchronized updating perioas the map content can be differ in geometry, accuracy and actuality. Geometric integration based on overlay result between PBB parcel map and recent...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/8266 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | ABSTRACT: <br />
<br />
<br />
<br />
<br />
Due to different acquisition method and not synchronized updating perioas the map content can be differ in geometry, accuracy and actuality. Geometric integration based on overlay result between PBB parcel map and recent Quickbird image show that the map layers no longer fit the new data. Due to misalignment pattern and inhomogeneous distribution of point displacement, the problem cannot be solved only by global transformation. <br />
<br />
<br />
<br />
<br />
The approach to improve data quality by applying best fitting' method using Iterative Closest Point Algorithm where common points determined by closest distance criteria between corresponding objects. <br />
<br />
<br />
<br />
<br />
Transformation using ICP Algorithm give the results that improving PBB parcel map quality can be achieved by improving RMS of global transformation result relative to segmented reference from average of 2,34 m to 1,14 m after applying ICP Algorithm. Interpolation method using Thin Plate Spline then applied to geometric smoothing and can improve RMS be average of 0,42 m: While RMS resulted from displacement to un-common points (check points on sample parcel) after applying ICP algorithm has average of 2,68 m, then applying Thin Plate Spline give RMS result 2,47 m in average. Area deviation from tolerance 10% to sample parcel area which assumed as non error system, on ICP Algorithm with Thin Plate Spline interpolation is equal to 32,6% while on transformation for initial approach and use of ICP Algorithm each equal to 43%. Disadvantages of correspondence points determination based on closest Euclidian distance is starting position should be close enough. |
|---|