2.
PROJECT BACKGROUND / INTRODUCTION - Brief on Ganoderma: A deadly
fungal disease threatening oil palm productivity. - Importance to FELCRA: High
infection risk and potential economic loss. - Focus Area: FELCRA plantations in Perak
(insert estate/block details). - Problem Statement: Late detection = late intervention
= yield loss.
3. OBJECTIVES - Detect Ganoderma infection before visible symptoms. - Use
drone-based imaging and AI/remote sensing for detection. - Produce risk maps for
early action. - Improve estate management response time.
4. METHODOLOGY Workflow/Steps: 1. Drone Data Collection
(multispectral/thermal/visual imagery) 2. Ground Truthing (field verification of
infected trees) 3. Image Processing (e.g. NDVI, thermal anomalies) 4. AI
Classification Model (optional: supervised ML) 5. Mapping Risk Zones via GIS 6.
Report Generation & Dashboard Upload
5. TOOLS & TECHNOLOGY USED - Drones: DJI Phantom 4 RTK / M350 + L2 Sensor
- Sensors: Multispectral / Thermal / RGB - Software: Pix4D, QGIS, ArcGIS Pro, Python
ML Toolkit - Data Layer: Tree point mapping, disease indicators, terrain overlays
6. PRELIMINARY FINDINGS - X hectares scanned, X suspected trees detected -
Heatmaps of infection likelihood - Zone classification: High / Medium / Low risk -
Sample image: infected vs healthy canopy pattern
7. CHALLENGES ENCOUNTERED - Dense canopy cover reduces detection
accuracy - Unstable weather disrupts flight schedule - Limited ground truth samples
- Differentiating Ganoderma from other stress factors
8. RECOMMENDATIONS - Confirmed zones: initiate isolation, felling, trenching -
Training for field staff to assist with mapping validation - Scale project to other
blocks or FELCRA estates - Integrate model into plantation dashboard system
9. IMPACT TO FELCRA - Early detection improves response time - Potential cost
saving from early intervention - Helps achieve better RSPO/MPOB compliance -
Strengthens estate’s disease management SOP
10. CONCLUSION & WAY FORWARD - Summary of tech effectiveness and results
- Expand deployment in 2025 (Phase 2) - Suggested monitoring cycle (e.g. every 3–
6 months) - Roadmap for integration with estate management system
11. APPENDIX / BACKUP SLIDES - Flight log examples - AI model confusion
matrix (if any) - Field photos of confirmed infected palms - GIS layer examples - Raw
data/Excel summary table
12. SPECIAL CASE STUDY: ADVANCED MULTISPECTRAL SIGNATURE
DETECTION
Spectral-Based Detection Method - Research shows Ganoderma infection alters
spectral reflectance, especially in Red, Red Edge, and NIR bands. - Healthy palms
(G0) reflect high NIR values; infected palms (G1–G3) reflect less NIR and more Red
due to stress.
Method Used with Drone: - Drone: DJI M350 with Multispectral Sensor
(e.g. Micasense RedEdge or Zenmuse L2) - Captured Bands: Blue, Green, Red, Red
Edge, NIR - Index Used: NDVI, NDRE, SAVI
Workflow: 1. Flight conducted over plantation blocks 2. Multispectral images
captured and stitched 3. Index maps generated to highlight stressed vegetation 4.
Threshold analysis or AI model classifies trees into G0, G1, G2, G3 5. Risk maps
created and compared with ground data
Result Expected: - Heatmap showing risk zones - Early detection before
symptoms - Accurate mapping of infected clusters - Quantified % of infected vs
healthy trees
Explanation to Stakeholders: “Based on spectral analysis, trees with reduced
NIR and Red Edge reflectance were identified as high-risk Ganoderma cases. This
allows proactive intervention before foliar symptoms appear, supporting early
felling or isolation and reducing spread.”
🔬 METHOD OVERVIEW: GANODERMA DETECTION USING MULTISPECTRAL DRONE
SENSORS
1. Data Acquisition
Drone Platform: DJI M350 RTK or equivalent
Sensor: Advanced Multispectral Camera (e.g., RedEdge-MX, Zenmuse L2)
Bands Captured:
Blue, Green, Red
Red Edge (very important for stress detection)
Near Infrared (NIR)
Why these bands?
→ Ganoderma affects water content & chlorophyll → this changes reflectance in Red
Edge & NIR bands before physical symptoms appear.
2. Spectral Signature Analysis
From the study you linked:
Healthy palms (G0) show higher reflectance in NIR, due to good chlorophyll &
internal water structure.
Infected palms (G1–G3) show:
Reduced NIR reflectance
Increased Red reflectance (due to chlorosis/stress)
Slight change in Red Edge
So the trick is: detect the shift in spectral curve before symptoms are visible.
3. Index Calculation
Use spectral indices that highlight vegetation stress:
Index Formula Function
NDVI (NIR – Red) / (NIR + Red) Detects general vegetation health
NDRE (NIR – Red Edge) / (NIR + Red Edge) More sensitive to early stress
SAVI ((NIR – Red) / (NIR + Red + L)) × (1 + L) Corrects for soil background
Use these indexes to generate:
Vegetation Stress Maps
Anomaly Zones
4. Classification / Thresholding
Method 1: Rule-Based Classification
Set threshold values based on:
NDVI < 0.4 = Possible Gano
NDRE < 0.25 = Moderate stress
Method 2: Supervised Machine Learning
Label training points: G0 (healthy), G1–G3 (infected)
Use classifier like:
Random Forest
SVM
KNN
Generate:
Risk Maps: classify trees as Healthy, Suspected, Infected
Tree Count Affected: quantification of each class
🎯 EXPECTED RESULT
Output Products:
High-resolution risk map: Shows suspected infected zones.
Tree health map: Color-coded per infection level.
Excel summary: SPH, infected %, area affected.
Time-series analysis (if multi-date flights are done): Disease spread trend.
