boc/iom/visual_geolocation/pipeline.py

"""
Visual Geolocation Pipeline
Complete pipeline from image to precise geolocation
"""

import sys
sys.path.insert(0, '/home/bernt/.openclaw/workspace/iom')

from typing import Dict, List, Optional, Tuple
from dataclasses import dataclass
from datetime import datetime
import numpy as np

from visual_geolocation.evidence_extractor import EvidenceExtractor, EvidencePackage
from visual_geolocation.geolocation_engine import GeolocationEngine, GeolocationEstimate
from visual_geolocation.confidence_model import ConfidenceModel, ConfidenceReport
from visual_geolocation.ocr_pipeline import OCRPipeline
from visual_geolocation.map_matcher import MapMatcher
from visual_geolocation.similarity_search import SimilaritySearch


@dataclass
class GeolocationResult:
    """Complete geolocation result"""
    lat: float
    lng: float
    accuracy: float
    confidence: float
    method: str
    evidence_summary: Dict
    confidence_report: ConfidenceReport
    map_matches: List[Dict]
    similar_images: List[Dict]


class VisualGeolocationPipeline:
    """
    Complete visual geolocation pipeline
    
    Flow:
    1. Extract evidence (7 layers)
    2. Geolocate (8 methods)
    3. Calculate confidence
    4. Match against maps
    5. Find similar images
    6. Return complete result
    """
    
    def __init__(self, use_real_ai: bool = True):
        self.evidence_extractor = EvidenceExtractor(use_real_ai=use_real_ai)
        self.geolocation_engine = GeolocationEngine()
        self.confidence_model = ConfidenceModel()
        self.ocr_pipeline = OCRPipeline(engine="simulated")
        self.map_matcher = MapMatcher()
        self.similarity_search = SimilaritySearch()
    
    def process_image(
        self,
        image_path: str,
        image_id: Optional[str] = None
    ) -> GeolocationResult:
        """
        Process image through complete pipeline
        
        Returns precise geolocation with confidence
        """
        print(f"Processing image: {image_path}")
        print("=" * 60)
        
        # Step 1: Extract evidence
        print("\n[1/6] Extracting evidence...")
        evidence = self.evidence_extractor.extract_all_evidence(image_path, image_id)
        
        # Step 2: Geolocate
        print("\n[2/6] Geolocating...")
        estimate = self.geolocation_engine.geolocate(evidence)
        
        # Step 3: Calculate confidence
        print("\n[3/6] Calculating confidence...")
        confidence_report = self.confidence_model.calculate_confidence(estimate, evidence)
        
        # Step 4: Match against maps
        print("\n[4/6] Matching against maps...")
        map_matches = self.map_matcher.match_location(
            estimate.lat,
            estimate.lng,
            evidence.to_dict(),
            radius=estimate.accuracy * 2
        )
        
        # Step 5: Find similar images
        print("\n[5/6] Finding similar images...")
        similar_images = []
        if evidence.visual_embedding is not None:
            similar_images = self.similarity_search.search(
                np.array(evidence.visual_embedding),
                k=5
            )
        
        # Step 6: Compile result
        print("\n[6/6] Compiling result...")
        result = GeolocationResult(
            lat=estimate.lat,
            lng=estimate.lng,
            accuracy=estimate.accuracy,
            confidence=confidence_report.overall_confidence,
            method=estimate.method,
            evidence_summary={
                "visual_objects": len(evidence.visual_objects),
                "semantic_objects": len(evidence.semantic_objects),
                "text_detections": len(evidence.text_detections),
                "geometric_features": len(evidence.geometric_features),
                "environmental_signals": len(evidence.environmental_signals)
            },
            confidence_report=confidence_report,
            map_matches=[
                {
                    "name": m.name,
                    "type": m.match_type,
                    "distance": m.distance,
                    "confidence": m.confidence
                }
                for m in map_matches[:5]
            ],
            similar_images=[
                {
                    "image_id": m.image_id,
                    "similarity": m.similarity,
                    "timestamp": m.timestamp
                }
                for m in similar_images[:5]
            ]
        )
        
        print("\n" + "=" * 60)
        print("PIPELINE COMPLETE")
        print("=" * 60)
        
        return result
    
    def print_result(self, result: GeolocationResult):
        """Print geolocation result"""
        print(f"\n📍 GEOLOCATION RESULT")
        print(f"   Position: {result.lat:.6f}, {result.lng:.6f}")
        print(f"   Accuracy: ±{result.accuracy:.1f}m")
        print(f"   Confidence: {result.confidence:.1%}")
        print(f"   Method: {result.method}")
        
        print(f"\n📊 EVIDENCE SUMMARY")
        for key, value in result.evidence_summary.items():
            print(f"   {key}: {value}")
        
        print(f"\n✅ CONFIDENCE REPORT")
        print(f"   Overall: {result.confidence_report.overall_confidence:.1%}")
        print(f"   Uncertainty: ±{result.confidence_report.uncertainty_radius:.1f}m")
        print(f"   Supporting evidence: {len(result.confidence_report.supporting_evidence)}")
        print(f"   Contradicting evidence: {len(result.confidence_report.contradicting_evidence)}")
        print(f"   Unknown factors: {len(result.confidence_report.unknown_factors)}")
        
        if result.map_matches:
            print(f"\n🗺️  MAP MATCHES")
            for match in result.map_matches:
                print(f"   {match['name']} ({match['type']})")
                print(f"   Distance: {match['distance']:.1f}m, Confidence: {match['confidence']:.1%}")
        
        if result.similar_images:
            print(f"\n🖼️  SIMILAR IMAGES")
            for img in result.similar_images:
                print(f"   {img['image_id']}: {img['similarity']:.1%} similarity")


# Example usage
def test_pipeline():
    """Test complete pipeline"""
    print("=" * 60)
    print("VISUAL GEOLOCATION PIPELINE")
    print("=" * 60)
    
    pipeline = VisualGeolocationPipeline(use_real_ai=True)
    
    # Create test image
    from PIL import Image
    img = Image.new('RGB', (640, 480), color='blue')
    img.save('/tmp/test_pipeline.jpg')
    
    # Process image
    result = pipeline.process_image('/tmp/test_pipeline.jpg', 'test_001')
    
    # Print result
    pipeline.print_result(result)
    
    return result


if __name__ == '__main__':
    test_pipeline()