import ee

# Initialize Earth Engine
ee.Initialize()

# Define the farm boundary as a GeoJSON geometry
farm_boundary = ee.Geometry.Polygon(
    [
				[
					[75.3589059, 24.4474494],
					[75.3588523, 24.4469977],
					[75.3594504, 24.4469415],
					[75.3594465, 24.4468412],
					[75.3598378, 24.4468241],
					[75.3598374, 24.4467459],
					[75.3600919, 24.446758],
					[75.3602018, 24.4473412],
					[75.3589059, 24.4474494]
				]
			]
)

# Define the time range for satellite imagery
start_date = '2022-01-01'
end_date = '2022-12-31'

# Load satellite imagery (e.g., Sentinel-2)
image_collection = (
    ee.ImageCollection('COPERNICUS/S2')
    .filterBounds(farm_boundary)
    .filterDate(ee.Date(start_date), ee.Date(end_date))
    .sort('CLOUDY_PIXEL_PERCENTAGE')
    .first()
)

# Define a region of interest for classification
roi = farm_boundary

# Load the USDA NASS Cropland Data Layer (CDL)
cdl = ee.Image('USDA/NASS/CDL/2021')  # You can choose the year you need

# Select relevant bands from the CDL
cdl_bands = ['cropland']

# Sample the CDL at the farm boundary
cdl_sample = cdl.select(cdl_bands).sample(
    region=roi,
    scale=30,
    geometries=True,
)

# Define the CDL class property as 'crop_type'
cdl_sample = cdl_sample.rename(['crop_type'])

# Merge the CDL sample with your own training data (if available)
# Replace 'training_data' with your own FeatureCollection
training_data = training_data.merge(cdl_sample)

# Select relevant bands from the imagery
bands = ['B2', 'B3', 'B4', 'B5', 'B6', 'B7', 'B8', 'B8A', 'B11', 'B12']

# Sample the imagery at training data points
training = image_collection.select(bands).sampleRegions(
    collection=training_data,
    properties=['crop_type'],
    scale=10,
)

# Define a classifier (e.g., Random Forest)
classifier = ee.Classifier.randomForest(10).train(
    features=training,
    classProperty='crop_type',
    inputProperties=bands,
)

# Classify the image
classified = image_collection.select(bands).classify(classifier)

# Get the crop type prediction for the farm boundary
prediction = classified.reduceRegion(
    reducer=ee.Reducer.mode(),
    geometry=roi,
    scale=10,
)

# Print the predicted crop type
print('Predicted Crop Type:', prediction.get('classification'))