Accuracy of Counting Coconut Palms with VHR Image

Accuracy of Counting Coconut Palms with VHR Image Data KB. Kataebati GSD – Geoscience Division

Texture Information IKONOS Geo. Eye World. View-2

1 st Step - Palm Resource Mapping with VHR Satellite Images Using World View-2 resolution 50 cm n o i t a c i l A. Separation of coconut palm from p p other vegetation A B. Density stratification g n i C. Selection of Plots s n D. Digitising e S E. Counting of palms per hectare e t o m e • Field Work R

Density Classes • Scattered coconut: 25 - 50 palms/hectare • Semi Dense: 50 – 150 palms/hectare • Dense Coconut > 150 palms/hectare

Digitising

Counting Palms in Plots • Placing a dot on top of every visible palm • Digital overlay with grid and counting in Map. Info (GIS software) dots per grid cell • Transfer to Access

The number of palms is counted through a semi automatic process connecting Map. Info GIS software with Access Database software Access Database Map. Info (GIS) Transfer

Establishment and Identify Plot in the Field using GPS- Rover (recording) Plot line GPS - Base Station Communicating with Satellite

Examples of Overlaying GPS Plot over the Image Plot Established from Satellite Image Identified by GPS

Mismatch of Plot Outlay • GPS accuracy of 80 cm does not apply for way points • Solution → new image interpretation for field plots

Comparison Between Image interpretation and in Reality 40 35 30 25 20 Images 15 Field 10 5 0 DCO CP SCO

Result Images VS Reality • 100 plots investigated • Dense coconut, 70 plots 10. 6% underestimated • Semi dense coconut 14 plots 4. 4% underestimated • Scattered coconut plots 16 plots correct image interpretation

Dense Coconut Palm

Conclusion • Factor 10. 6% and 4. 4% is incorporated in code to calculate palms per hectare • Accuracy will be improved with additional investigations

Thanks