Geoscripting project repository

• Title: Fantastic Palms: The Secret Hidden in Planet Images
• Team name： lovely palm trees
• Challenge number: 10

Research Question

Mauritia flexuosa Extraction in Aguajales This model is designed to extract Mauritia flexuosa in the aguajales training region and predict its density and distribution in the prediction area.

Steps to Achieve the Objective

1. Data Retrieval

Download and extract the raw data.

2. Data Preparation

Clip the data to define the training and prediction areas.

3. Object Segmentation

Utilize the segment-anything model to generate masks for all objects in the UAV image for both the training and prediction areas.

4. Raster-to-Vector Conversion

Convert the raster images into vector format.

5. Classification in Training Area

Classify and label all the polygons in the training area to prepare for the subsequent prediction step.

6. Prediction

Based on whether the segmentation from step 3 was performed:

• Based on Pixel: Utilize SVM (Support Vector Machine) and RF (Random Forest) for modeling and prediction.
• Based on Object: Similarly, utilize SVM and RF for modeling and prediction.

7. Visualization

Visualize the extracted and predicted Mauritia flexuosa density and distribution in both training and prediction areas.

Notes

Ensure that the appropriate libraries and dependencies for the above tools (like segment-anything, SVM, RF) are installed and set up correctly.

Result

Testing

To set up the project environment, create a virtual environment based on finalproject.yaml. Then installation dependencies:

mamba env create --file finalproject.yaml
source activate finalproject

execute the program

./main.sh

Packages Usage

Python Dependencies

• Download and extract modified_1:
  • os
  • requests
  • zipfile
• Clip_modified_2:
  • geopandas
  • rasterio
  • shapley
• Segment_modified_3:
  • pillow
  • segment_anything
  • numpy
  • matplotlib
  • requests
  • opencv-python
• Vectorize_modified_4:
  • gdal
  • ogr

R Dependencies

• Set_label_5:
  • raster
  • terra
  • sf
  • ranger
  • dplyr
• Model_of_prediction_6 (Python):
  • sklearn
• Visualization_leaflet_7:
  • leaflet
  • png
  • sp

License

The code is licensed under the MIT License.

Contributors

• Xinyi He
• Xiaoyu Yang
• Dong Liang
• Qin Xu