Project Overview

This document provides a comprehensive guide on how to process DJI P4 Multispectral (P4M) multispectral imagery using open-source software WebODM and QGIS, from environment setup to final vegetation index extraction.

Complete Workflow for DJI P4M Multispectral Image Processing with WebODM and QGIS

Part 1: Environment Setup and Software Installation

1. Software Downloads

Ensure the following base software is installed:

• Git: For cloning code repositories.
• Docker Desktop: The container environment required for WebODM.
• Python (v3.0+): Script execution environment.

2. WebODM Installation and Deployment

Mac / Windows (Bash)

Open Terminal (Terminal or Git Bash) and execute the following commands:

cd D:/  # Or any other directory where you want to install
# Set Git proxy (optional, if network is restricted)
# git config --global http.proxy 'http://proxy_ip:port'
# git config --global https.proxy 'http://proxy_ip:port'

# Clone WebODM repository
git clone https://github.com/OpenDroneMap/WebODM
cd WebODM

# Docker login (optional, if you need to pull private images or avoid rate limits)
# docker login -u <username>

# Start WebODM
./webodm.sh start

3. Docker Image Acceleration (Recommended for China)

If download speeds are slow, you can add mirror accelerators in Docker Desktop settings:

"registry-mirror": [
  "https://docker.mirrors.ustc.edu.cn",
  "https://mirror.ccs.tencentyun.com"
]

Restart WebODM:

cd WebODM
./webodm.sh stop
./webodm.sh start

4. Accessing WebODM

1. Open browser and visit http://localhost:8000.
2. Create username and password on first login.
3. Enter the console to start processing tasks.

---

Part 2: WebODM Image Processing

1. Multispectral Data Preparation (Critical)

Original Data Requirements (Must Follow):

• Must retain EXIF information: Especially DLS (Daylight Sensor) information for radiometric calibration.
• Do not rename files: Keep original filenames.
• P4M Data Structure: Each capture point should contain 6 images:
  • RGB (jpg)
  • Blue (tif)
  • Green (tif)
  • Red (tif)
  • RedEdge (tif)
  • NIR (tif)

Important Note: WebODM automatically parses bands, so there's no need to manually separate band files. Just upload all images directly.

2. Creating Projects and Tasks

1. New Project: Click New Project, name it (e.g.): maize_N_trial_2025.
2. New Task: Click New Task, upload all multispectral images (WebODM will automatically parse bands without manual classification).

3. Processing Parameter Settings (Core Steps)

Click Task Options, use the following settings to ensure multispectral data accuracy:

Recommended Preset: Select Multispectral preset, or manually confirm the following options.

Must Check/Enable:

• Radiometric calibration: Radiometric correction (using DLS information).
• Use EXIF GPS: Use GPS positioning.
• Use fixed camera parameters: Fixed camera parameters.
• Use band alignment: Band registration (ensure spatial alignment across bands).

Disable Unnecessary Modules (Save Time):

• Skip 3D model: Skip 3D model generation.
• Skip point cloud: Skip point cloud generation.
• Skip textured model: Skip textured model.

Resolution and Quality:

• Feature quality: Medium (balance speed and quality).
• Orthophoto resolution: Leave blank (automatic, recommended to keep original resolution).

Click Start Task to begin processing.

4. Exporting Results

After task completion, go to Assets -> Results to download results. Core file: orthophoto.tif (multiband multispectral orthophoto mosaic).

---

Part 3: QGIS Post-processing and Analysis

1. Data Loading and CRS Check

1. Drag orthophoto.tif into QGIS.
2. Check Coordinate System (CRS):
   • Right-click layer -> Properties -> Information.
   • Confirm CRS is a projected coordinate system (e.g., EPSG:32645 WGS84 / UTM Zone 45N), units should be meters.
   • Note: All subsequent vector layers must match this coordinate system.

2. Clipping Field Area (Clip)

1. Create Boundary: If no boundary file exists, create a new GeoPackage layer (Layer -> Create Layer -> New GeoPackage Layer), type Polygon, consistent CRS. Manually draw field boundaries and save.
2. Clip by Mask:
   • Toolbox path: GDAL -> Raster extraction -> Clip raster by mask layer.
   • Input raster: Orthophoto.
   • Mask layer: Field boundary Polygon.
   • Check Crop to cutline.
   • Output: orthophoto_field.tif.

3. Calculating Vegetation Index (NDVI)

Assume band order: Red (Band 1), NIR (Band 4) （Please confirm P4M band order according to camera manual, as P4M Tif output order may vary, check band properties first）.

Formula: NDVI = (NIR - Red) / (NIR + Red)

1. Open Raster -> Raster Calculator.
2. Input formula (example):

   ("xkx-1-orthophoto@4" - "xkx-1-orthophoto@1") / ("xkx-1-orthophoto@4" + "xkx-1-orthophoto@1")

3. Output file: NDVI.tif.

4. Generating Vegetation Mask (Removing Soil Background)

To ensure accurate statistics, remove interference from bare soil between rows.

1. Open Raster Calculator.
2. Formula:

   ("NDVI@1" > 0.2) * "NDVI@1"

   Explanation: When NDVI > 0.2, retain original value, otherwise set to 0 (or NoData).
3. Output: NDVI_veg.tif.

5. Grayscale Quantization (GLCM Texture Calculation Preparation)

GLCM texture calculation typically requires integer grayscale images, while NDVI is floating-point.

1. Open Raster Calculator.
2. Formula: Map -11 NDVI to 0255 integers.

   ("NDVI_veg@1" + 1) * 127

3. Critical Setting: Output data type must select Byte (UInt8).
4. Output: NDVI_veg_uint8.tif.

6. Calculating GLCM Texture Features

Important Prerequisite: Grayscale Quantization GLCM texture calculation requires input images to have integer grayscale values, while NDVI is Float32 continuous values (range -1~1), direct calculation will cause errors. Grayscale quantization is mandatory:

1. Grayscale Quantization Steps:

   • Open Raster -> Raster Calculator
   • Formula: ("NDVI_veg@1" + 1) * 127
   • Critical Setting: Output data type must select Byte (UInt8)
   • Output: NDVI_veg_uint8.tif
   • Verification: Right-click layer -> Properties -> Information, confirm Data type = Byte

2. GLCM Texture Calculation:

   • Toolbox path: GRASS -> Raster (r.*) -> r.texture
   • Input raster: NDVI_veg_uint8.tif
   • Neighborhood size (Window size):
     • UAV 5-10cm resolution: Recommended 3 or 7
     • Larger windows = smoother texture features; smaller windows = richer details
   • Check texture metrics:
     • asm (Angular Second Moment)
     • contrast (Contrast)
     • corr (Correlation)
     • var (Variance)
     • idm (Inverse Difference Moment)
     • entr (Entropy)
     • dv (Variance)
     • sa (Angular Second Moment)
   • Output prefix: NDVI_tex
   • Generated files:
     • NDVI_tex_asm.tif
     • NDVI_tex_contrast.tif
     • NDVI_tex_corr.tif
     • NDVI_tex_var.tif
     • NDVI_tex_idm.tif
     • NDVI_tex_entr.tif
     • NDVI_tex_dv.tif
     • NDVI_tex_sa.tif

7. Plot-scale Feature Extraction (Zonal Statistics)

Finally, aggregate raster pixel values to each experimental plot (Plot).

1. Toolbox path: Raster analysis -> Zonal statistics.
2. Raster layer: NDVI_veg.tif (or other texture layers).
3. Vector layer containing zones: Plot polygon layer (Polygon).
4. Zone field: Plot ID (plot_id).
5. Statistics: Select Mean (mean), StDev (standard deviation), etc.
6. Output column prefix: e.g., NDVI_ZS_ (Zonal Statistics abbreviation).

After running, the plot layer's attribute table will contain calculated statistics:

• NDVI_ZS_mean: Plot NDVI mean
• NDVI_ZS_std: Plot NDVI standard deviation

You will eventually obtain a plot-level feature table that can be exported as CSV for subsequent agricultural statistical analysis.

---

Important Notes and Best Practices

Data Preparation Phase

1. Data Integrity: Retain all EXIF information from original TIFF files, especially DLS data, which is the foundation for radiometric calibration.
2. File Naming: Never rename original files, keep camera-generated filenames.
3. Data Integrity Check: Before uploading, confirm each capture point contains all 6 band files.

WebODM Processing Phase

1. Parameter Selection: The Multispectral preset includes most necessary parameters, but manually confirm radiometric calibration and band alignment are enabled.
2. Resolution Settings: Leave Orthophoto resolution blank to maintain original resolution, or specify specific values (e.g., 5cm) if downsampling is needed.
3. Time Control: Disabling 3D-related modules significantly reduces processing time.

QGIS Post-processing Phase

1. CRS Consistency: All subsequent analyses must be performed in a unified projected coordinate system to avoid area calculation errors.
2. Band Order Confirmation: P4M band order may vary with firmware version, always check band properties before calculating indices.
3. Mask Threshold Selection: NDVI threshold 0.2 is empirical, adjust based on actual crop types and soil background.
4. GLCM Window Size: Window size affects the spatial scale of texture features, choose based on research objectives and UAV resolution.

Result Validation

1. Intermediate Result Check: Visualize results after each step to ensure no obvious errors.
2. Data Type Verification: Confirm input is Byte type before GLCM calculation.
3. Statistical Reasonableness: Check final plot statistics are within reasonable ranges (e.g., NDVI typically 0.2-0.8).