LANDROID

AI-Powered Land Intelligence Platform

Hackathon Prototype — Birdscale Technology × VIT Chennai | April 2026

Overview

Landroid is an AI-powered mobile-first platform that transforms raw drone survey data into actionable land intelligence. Built for land consultants and landowners, it provides real-time land health monitoring, plant intelligence, tree canopy analysis, and AI-driven land valuation — all on a single interactive GIS map.

The platform ingests Birdscale-provided geospatial datasets (orthomosaics, DEM, NDVI rasters, boundary GeoJSON) and enriches them with live data from 5 free open APIs to produce computed, confidence-scored AI outputs.

Key Highlights

3 fully functioning AI modules with dynamic confidence scores on every output
Real-time tile server rendering multi-gigabyte GeoTIFF rasters as on-demand Web Mercator tiles
Live API integrations — ISRIC SoilGrids, OSM Overpass, OSM Nominatim, Microsoft Planetary Computer, Open-Meteo
Two distinct user roles — Land Consultant (Admin) and Landowner (Read-only) with access control
English + Tamil language support

Architecture

┌──────────────────────────────────────────────────────────────┐
│                     Flutter Mobile App                        │
│  MapLibre GL  ·  Firebase Auth  ·  Provider State Mgmt       │
│  Glassmorphic UI  ·  Multi-language (EN / Tamil)             │
└──────────────────────┬───────────────────────────────────────┘
                       │ HTTP / REST
┌──────────────────────▼───────────────────────────────────────┐
│                    FastAPI Backend                            │
│                                                              │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐       │
│  │ Land Health  │  │ Plant Health │  │ Tree & Canopy│       │
│  │ Dashboard    │  │ Zone Map     │  │ Count        │       │
│  │ (FR-18→24)  │  │ (FR-25→28)  │  │ (FR-29→33)  │       │
│  └──────┬───────┘  └──────┬───────┘  └──────┬───────┘       │
│         │                 │                 │               │
│  ┌──────▼─────────────────▼─────────────────▼───────┐       │
│  │              Tile Server (z/x/y.png)              │       │
│  │  Windowed reads · Vectorised colormaps · Cache    │       │
│  └──────────────────────┬────────────────────────────┘       │
│                         │                                    │
│  ┌──────────────┐  ┌────▼─────────┐  ┌──────────────┐       │
│  │ Land         │  │ Geofencing   │  │ Document     │       │
│  │ Valuation    │  │ & Alerts     │  │ Vault        │       │
│  │ (FR-34→38)  │  │ (FR-39→43)  │  │ (FR-44→47)  │       │
│  └──────────────┘  └──────────────┘  └──────────────┘       │
└──────────────────────────────────────────────────────────────┘
         │                    │                │
    ┌────▼────┐         ┌────▼────┐      ┌────▼────┐
    │ SoilGrids│        │ OSM     │      │Planetary│
    │ (ISRIC)  │        │Overpass │      │Computer │
    │          │        │Nominatim│      │Open-Meteo│
    └──────────┘        └─────────┘      └──────────┘

AI Modules

Module 1: Land Health Dashboard (FR-18 → FR-24)

Aggregates four environmental signals into a single composite Land Health Score (0–100).

Signal	Weight	Data Source	What It Measures
NDVI Trend	40%	Birdscale NDVI raster + Sentinel-2	Vegetation health & historical trend
Rainfall Adequacy	30%	Open-Meteo Historical API	Annual/monthly precipitation vs. regional normal
Soil Quality	20%	ISRIC SoilGrids REST API	pH, organic carbon, texture classification
Temperature Suitability	10%	Open-Meteo Historical API	Monthly trend, heat stress event count

Score Mapping: Healthy (75–100) · Moderate (50–74) · At Risk (below 50)

Each signal card displays: current value, trend indicator, and a per-signal confidence score.

Module 2: Plant Health Zone Map (FR-25 → FR-28)

Classifies the NDVI raster into four health zones rendered as a colour-coded overlay:

Zone	NDVI Range	Colour
Bare/Stressed	< 0.2	Brown
Sparse	0.2 – 0.4	Amber
Healthy	0.4 – 0.6	Light Green
Dense	> 0.6	Dark Green

Computes percentage area in each zone with dynamic confidence based on valid pixel ratio and NDVI spread.

Module 3: Tree & Canopy Count (FR-29 → FR-33)

OpenCV watershed-based pipeline for individual tree canopy detection:

Orthomosaic → Excess Green Index (2G-R-B) → Gaussian Blur → Otsu Threshold
→ Morphological Opening → Distance Transform → Watershed Segmentation
→ Contour Detection → Canopy Count + Stress Flagging

Outputs:

Total canopy count and density per acre
GeoJSON FeatureCollection of canopy centroids overlaid on the orthomosaic
Stressed canopy detection (area < 50% of parcel median)
Cross-date comparison: new/missing canopies flagged with visual markers
Confidence score derived from GSD (ground sampling distance) and detection clarity

Module 4: Land Valuation (FR-34 → FR-38)

Estimates land value (Rs. per acre) using five weighted signals:

Signal	Weight	Source
Land Health Score	30%	Module 1 output
OSM Proximity (highway, town, water)	25%	OSM Overpass API
Soil Quality	20%	ISRIC SoilGrids API
Rainfall Adequacy	15%	Open-Meteo / CHIRPS
Night Light Index	10%	Planetary Computer VIIRS

Output: Low / Mid / High value band with Rs./acre range, confidence score (%), and top 3 driving factors listed with positive/negative impact.

All valuation outputs are labelled as estimated intelligence ranges, not legal or government guideline valuations.

Open APIs Used

All APIs are free with no credit card required.

API	What It Provides	Signup	Used By
ISRIC SoilGrids REST API	Soil type, pH, organic carbon, texture by coordinates	None — immediate	Health Dashboard, Valuation
OSM Overpass API	Highway, water body, town proximity by bounding box	None — immediate	Valuation (proximity signals)
OSM Nominatim	Reverse geocoding — town, taluk, district, state	None — immediate	Valuation (location context)
Microsoft Planetary Computer	VIIRS night-light radiance, Sentinel-2 NDVI time series	Free email signup	Valuation, NDVI history
Open-Meteo Historical Weather	Monthly rainfall (CHIRPS-equivalent), temperature (ERA5-equivalent)	None — immediate	Health Dashboard (climate signals)

API Flow

When a parcel boundary is confirmed, the backend derives the centroid and bounding box and fires all API calls in parallel using asyncio.gather():

Boundary GeoJSON → Centroid + BBox
    ├── ISRIC SoilGrids    (centroid)     → pH, organic carbon, texture
    ├── OSM Overpass        (bbox + 5km)  → highway/water/town proximity
    ├── OSM Nominatim       (centroid)    → town, district, state
    ├── Planetary Computer  (centroid)    → VIIRS night-light radiance
    └── Open-Meteo Archive  (centroid)    → rainfall + temperature history

Tech Stack

Layer	Technology
Mobile Framework	Flutter (Dart)
Map Rendering	MapLibre GL (`maplibre_gl` package)
Backend Framework	FastAPI (Python) with Uvicorn
Geospatial Processing	Rasterio, PyProj, Shapely
Computer Vision	OpenCV (watershed segmentation)
ML / Data	NumPy, scikit-learn (TrendPredictor)
Authentication	Firebase Auth (OTP + Google Sign-In)
State Management	Provider
Tile Serving	Custom real-time tile renderer (windowed reads, vectorised colormaps, in-memory cache)

Repository Structure

landroid/
├── backend/                      # FastAPI Backend
│   ├── main.py                   # App entrypoint, router registration, CORS
│   ├── routers/                  # API endpoint handlers
│   │   ├── health_dashboard.py   # Land Health Dashboard (Module 1)
│   │   ├── plant_zones.py        # Plant Health Zone Map (Module 2)
│   │   ├── canopy_count.py       # Tree & Canopy Count (Module 3)
│   │   ├── valuation.py          # Land Valuation (Module 4)
│   │   ├── tiles.py              # Real-time tile server (NDVI, DEM, Zones, Terrain)
│   │   ├── boundary.py           # Boundary GeoJSON + centroid/bbox
│   │   ├── documents.py          # Document Vault CRUD
│   │   ├── geofencing.py         # Geofence rules and alerts
│   │   ├── time_series.py        # NDVI time series data
│   │   └── trends.py             # ML trend predictions
│   ├── services/                 # Core business logic & API clients
│   │   ├── soilgrids.py          # ISRIC SoilGrids REST API client
│   │   ├── osm.py                # OSM Overpass + Nominatim client
│   │   ├── planetary.py          # Planetary Computer VIIRS + Open-Meteo climate
│   │   ├── canopy.py             # OpenCV watershed canopy detection pipeline
│   │   ├── ndvi_service.py       # Sentinel-2 NDVI time series fetcher
│   │   ├── scoring.py            # Land valuation weighted formula
│   │   ├── parcel_registry.py    # Parcel data management
│   │   └── time_series.py        # Time series processing
│   ├── ml/
│   │   └── trend_predictor.py    # TrendPredictor ML model
│   ├── models/
│   │   └── schemas.py            # Pydantic request/response schemas
│   ├── utils/
│   │   ├── confidence.py         # Confidence score calculations for all modules
│   │   ├── geo.py                # Haversine distance and geo utilities
│   │   └── raster.py             # GeoTIFF download, read, and cleanup
│   └── tests/                    # Pytest test suites
├── mobile/                       # Flutter Mobile App
│   └── lib/
│       ├── main.dart             # App entry point
│       ├── core/                 # Constants, theme, localizations (EN + Tamil)
│       ├── models/               # Dart data models (canopy, health, valuation, zones)
│       ├── providers/            # State management (language provider)
│       ├── screens/              # UI screens (map, dashboard, auth, insights, etc.)
│       ├── services/             # API service + Auth service
│       └── widgets/              # Reusable UI (GlassCard, HealthRing, MetricCard)
├── data/                         # (Gitignored) Birdscale raster data & GeoJSONs
├── docs/                         # Architecture, API contracts, ML overview
├── .env.example                  # Environment variable template
├── .gitignore                    # Ignores .env, data/*, node_modules
└── requirements.txt              # Python dependencies

Getting Started

Prerequisites

Python 3.10+
Flutter SDK 3.11+
Android Studio / Emulator (API 29+)
Git

1. Clone the Repository

git clone <repository-url>
cd landroid

2. Backend Setup

cd backend

# Create and activate virtual environment
python -m venv .venv
# Windows:
.venv\Scripts\activate
# macOS/Linux:
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Configure environment
cp ../.env.example .env
# Edit .env — fill in PLANETARY_COMPUTER_SUBSCRIPTION_KEY if available

# Start the server
uvicorn main:app --host 0.0.0.0 --port 8000 --reload

Important — Warm up tile cache for instant map loading:

GET http://localhost:8000/api/v1/tiles/warmup

Interactive API docs: http://localhost:8000/docs

3. Place Birdscale Data

Place the provided geospatial files in the data/ directory (these are gitignored and never committed):

data/
├── Orthomosaic.tif               # Drone orthomosaic (GeoTIFF)
├── ndvi.tif                      # NDVI raster (GeoTIFF)
├── Digital Elevation model.tif   # DEM elevation raster (GeoTIFF)
└── boundary_wgs84.geojson        # Parcel boundary (GeoJSON)

4. Mobile Setup

cd mobile

# Install Flutter dependencies
flutter pub get

# Update API base URL in lib/core/app_constants.dart
# Set it to your machine's local IP (e.g., http://192.168.x.x:8000)

# Run on device or emulator
flutter run

API Endpoints

All endpoints are prefixed with /api/v1.

Method	Endpoint	Description
`POST`	`/health-dashboard`	Land Health Dashboard — composite score + 4 signals
`POST`	`/plant-zones`	Plant Health Zone classification (4 NDVI categories)
`POST`	`/canopy-count`	Tree & canopy detection with GeoJSON overlay
`POST`	`/valuation`	Land valuation estimate with factor breakdown
`GET`	`/tiles/ndvi/{z}/{x}/{y}.png`	NDVI colour-ramp tile layer
`GET`	`/tiles/dem/{z}/{x}/{y}.png`	DEM elevation colour-ramp tile layer
`GET`	`/tiles/zones/{z}/{x}/{y}.png`	Plant health zone colour-coded tile layer
`GET`	`/tiles/terrain/{z}/{x}/{y}.png`	Mapbox Terrain-RGB encoded tiles
`GET`	`/tiles/warmup`	Pre-render all parcel tiles for instant loading
`POST`	`/boundary`	Parcel boundary GeoJSON, centroid, bbox, area
`POST`	`/documents/upload`	Upload document to vault (Land Consultant only)
`GET`	`/documents/{parcel_id}`	List documents for a parcel
`POST`	`/geofencing/rules`	Create geofence rule
`GET`	`/geofencing/{parcel_id}`	List geofence rules and alerts
`POST`	`/time-series`	NDVI time series data
`POST`	`/trends`	ML trend predictions
`GET`	`/health`	Server health check

Confidence Scoring

Every AI output includes a confidence score (0–100%) that reflects data quality, not just a static label.

How Confidence Is Computed

Canopy Count: Weighted combination of image resolution (GSD), canopy count plausibility, and null pixel ratio.

Land Health Dashboard: Based on NDVI raster valid pixel coverage, and whether soil (SoilGrids), rainfall (Open-Meteo), and temperature data came from live APIs vs. fallbacks.

Land Valuation: Starts at 60%, then adjusts based on whether OSM proximity signals were found (+15/-10), VIIRS night-light is real (+10/-8), SoilGrids responded (+8/-5), and rainfall data is live (+7/-5).

Plant Zones: Derived from NDVI valid pixel ratio and spectral spread of values.

User Roles

Feature	Land Consultant (Admin)	Landowner (User)
Create / edit parcels	Yes	No
Upload drone rasters	Yes	No
Draw boundary polygon	Yes	No
Assign parcel to Landowner	Yes	No
Land Health Dashboard	All parcels	Own parcels only
Plant Health Zone Map	All parcels	Own parcels only
Tree & Canopy Count	All parcels	Own parcels only
Land Valuation	All parcels	Own parcels only
Document Vault — upload	Yes	No
Document Vault — view	All parcels	Own parcels only
Geofencing alerts	All parcels	Own parcels only

Testing

cd backend
pytest tests/ -v

Test suites cover:

Canopy detection pipeline
Health dashboard scoring
Plant zone classification
Land valuation formula
API routing and error handling

Security

Zero credentials in source code — all API keys and tokens stored in .env (gitignored)
.gitignore excludes .env, .venv, data/*, __pycache__/
All Birdscale-provided datasets are gitignored and fetched at runtime
Bearer token authentication on API requests
Firebase Auth with Android Keystore for secure token storage
No permanent public document sharing links — time-limited only (48 hours)

Environment Variables

# .env.example
PLANETARY_COMPUTER_SUBSCRIPTION_KEY=   # Optional — public collections work without key
APP_ENV=development                     # development | production
LOG_LEVEL=INFO

Demo Narrative

"The Land Consultant created this parcel and assigned it to the Landowner. The Land Health Score is 58 — Moderate. NDVI has dropped 12% over 2 years and rainfall is 9% below normal. The NDVI zone map shows 34% of the parcel is in the stressed zone. We detected 386 trees; 11 are flagged as potentially stressed. Estimated value: Rs. 2.4L to Rs. 3.1L per acre."

All outputs are dynamically computed from real data — changing the input parcel changes every result.

Built for the Birdscale Technology × VIT Chennai Hackathon — April 2026

Navigation Menu

Repository files navigation

LANDROID

AI-Powered Land Intelligence Platform

Overview

Key Highlights

Architecture

AI Modules

Module 1: Land Health Dashboard (FR-18 → FR-24)

Module 2: Plant Health Zone Map (FR-25 → FR-28)

Module 3: Tree & Canopy Count (FR-29 → FR-33)

Module 4: Land Valuation (FR-34 → FR-38)

Open APIs Used

API Flow

Tech Stack

Repository Structure

Getting Started

Prerequisites

1. Clone the Repository

2. Backend Setup

3. Place Birdscale Data

4. Mobile Setup

API Endpoints

Confidence Scoring

How Confidence Is Computed

User Roles

Testing

Security

Environment Variables

Demo Narrative

About

Releases

Packages

Contributors

Languages

Name		Name	Last commit message	Last commit date
backend		backend
data		data
docs		docs
mobile		mobile
.env.example		.env.example
.gitignore		.gitignore
README.md		README.md
ip_info.txt		ip_info.txt
problems.md		problems.md
requirements.txt		requirements.txt

Latest commit

History

Folders and files

Repository files navigation

LANDROID

AI-Powered Land Intelligence Platform

Overview

Key Highlights

Architecture

AI Modules

Module 1: Land Health Dashboard (FR-18 → FR-24)

Module 2: Plant Health Zone Map (FR-25 → FR-28)

Module 3: Tree & Canopy Count (FR-29 → FR-33)

Module 4: Land Valuation (FR-34 → FR-38)

Open APIs Used

API Flow

Tech Stack

Repository Structure

Getting Started

Prerequisites

1. Clone the Repository

2. Backend Setup

3. Place Birdscale Data

4. Mobile Setup

API Endpoints

Confidence Scoring

How Confidence Is Computed

User Roles

Testing

Security

Environment Variables

Demo Narrative

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages