🌳 Object Detection and Segmentation of Trees using Text SAM in ArcGIS Online

🚀 This repository provides an English-only, step-by-step GeoAI workflow for tree object detection and segmentation from high-resolution aerial imagery using Text SAM (Segment Anything Model) within ArcGIS Online Notebooks and the ArcGIS API for Python.

🧠 The workflow outputs vectorized tree polygons, enabling advanced spatial analysis, urban greening assessment, and decision support in ArcGIS Pro.

---

📚 Related Publication (Official / Legitimate Sources)

• 📘 Esri Press — Book Page
  https://www.esri.com/en-us/esri-press/browse/security-first-geospatial-workflows-for-a-safe-and-equitable-world

• 📄 ResearchGate — Chapter Page
  https://www.researchgate.net/publication/396514864_Object_detection_and_segmentation_of_trees_using_Text_SAM_in_ArcGIS_Online

⚠️ Note
This repository supports learning, reproducibility, and teaching for the workflow described in the chapter.
Please consult the links above for the official and legitimate publication record.

---

🎯 What You Will Do

By following this workflow, you will:

1. 🛰️ Upload a high-resolution aerial image as an Imagery Layer in ArcGIS Online
2. ⚡ Launch an ArcGIS Online Notebook (Advanced with GPU support)
3. 🤖 Retrieve the Text SAM deep learning package hosted on ArcGIS Online
4. 🌲 Perform tree segmentation using a natural-language text prompt (e.g., "tree")
5. 🗺️ Export and analyze vector tree polygons in ArcGIS Pro

---

🧰 Requirements

🧭 ArcGIS / Esri

You will need:

• An ArcGIS Online account with access to:
  • ✅ ArcGIS Notebooks for ArcGIS Online
  • ✅ ArcGIS Image for ArcGIS Online (required for Imagery Layer publishing)

📦 Data

• High-resolution aerial imagery (GeoTIFF recommended)
• Example source: OpenAerialMap
• Chapter example: Estella Public School (New South Wales, Australia)

---

⚡ Quick Start (ArcGIS Online)

🟢 Step 1 — Upload imagery as an Imagery Layer

1. Download an aerial image (GeoTIFF recommended).
2. In ArcGIS Online: Content → New Item → Imagery Layer
3. Select:
   • Tiled Imagery Layer
   • One Image
4. Upload and publish the imagery.

---

🟢 Step 2 — Create a GPU Notebook

1. Navigate to Notebooks in ArcGIS Online.
2. Select New Notebook → Advanced with GPU support

   ⚠️ Note: GPU notebooks consume ArcGIS credits.

---

🟢 Step 3 — Run the workflow

📂 Open: ➡️ Copy each section into ArcGIS Notebook cells and run sequentially.

---

📤 Output

• 🌳 detectedTrees — Feature Layer (tree polygons)
• 👀 Viewable in ArcGIS Online Map Viewer
• 🧮 Importable into ArcGIS Pro for:
  • Area calculation
  • Density analysis
  • Proximity & accessibility studies
  • Canopy coverage estimation

---

🔍 Recommended Follow-up Analyses (ArcGIS Pro)

Once tree polygons are generated, consider:

• 📐 Tree canopy coverage (aggregate polygon area)
• 📊 Tree density surfaces (kernel density, tessellation summaries)
• 🚸 Proximity to infrastructure (schools, roads, public facilities)
• 🌡️ Equity & environmental exposure overlays
  (urban heat, air quality, vulnerable communities)

---

🎛️ Notes on Accuracy & Performance

Segmentation quality depends on:

• 🖼️ Imagery resolution
• 🧩 Scene complexity
• 📝 Text prompt specificity

Recommended tuning strategies:

• Adjust box / text thresholds
• Modify padding and batch size
• Apply post-processing:
  • Eliminate sliver polygons
  • Dissolve adjacent features
  • Simplify geometry
  • Run topology checks

---

📝 Citation

If you use or extend this workflow in academic work, please cite:

• The corresponding book chapter
• Relevant SAM / Text SAM references

---

📜 License

🔓 Add a license appropriate to your use case (e.g., MIT License for code).

⚠️ If figures or text from the book chapter are reused, ensure compliance with Esri Press publisher permissions.

---

📬 Contact

If you encounter:

• Missing code segments
• Reproducibility issues
• Questions about extending the workflow

📧 Please contact:
Yifan Yang
(see email in GitHub profile or publication contact information)