boc/iom/visual_geolocation/temporal_analysis.py

"""
Temporal Analysis
Compare multiple observations of the same location over time
"""

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 (
    EvidencePackage, ImageMetadata, VisualObject, 
    TextDetection, GeometricFeature, EnvironmentalSignal
)


@dataclass
class TemporalComparison:
    """Comparison between two observations"""
    observation_id_1: str
    observation_id_2: str
    time_delta: float  # hours
    
    # Changes detected
    new_objects: List[Dict]
    removed_objects: List[Dict]
    changed_objects: List[Dict]
    
    # Stability metrics
    structural_stability: float  # 0-1
    activity_change: float  # 0-1
    lighting_change: float  # 0-1
    
    # Reality gap evolution
    rgi_delta: float  # Change in RGI


class TemporalAnalyzer:
    """
    Analyze temporal changes between observations
    
    Use cases:
    - Detect infrastructure degradation
    - Monitor construction progress
    - Track urban changes
    - Validate maintenance effectiveness
    """
    
    def __init__(self):
        self.observations: Dict[str, EvidencePackage] = {}
    
    def add_observation(self, observation_id: str, evidence: EvidencePackage):
        """Add observation to temporal database"""
        self.observations[observation_id] = evidence
    
    def compare_observations(
        self,
        obs_id_1: str,
        obs_id_2: str
    ) -> TemporalComparison:
        """Compare two observations"""
        obs1 = self.observations[obs_id_1]
        obs2 = self.observations[obs_id_2]
        
        # Calculate time delta
        time_delta = self._calculate_time_delta(obs1, obs2)
        
        # Compare visual objects
        new_objects, removed_objects, changed_objects = self._compare_objects(
            obs1.visual_objects,
            obs2.visual_objects
        )
        
        # Calculate stability metrics
        structural_stability = self._calculate_structural_stability(
            obs1, obs2, new_objects, removed_objects
        )
        
        activity_change = self._calculate_activity_change(obs1, obs2)
        lighting_change = self._calculate_lighting_change(obs1, obs2)
        
        # Calculate RGI delta
        rgi_delta = self._calculate_rgi_delta(obs1, obs2)
        
        return TemporalComparison(
            observation_id_1=obs_id_1,
            observation_id_2=obs_id_2,
            time_delta=time_delta,
            new_objects=new_objects,
            removed_objects=removed_objects,
            changed_objects=changed_objects,
            structural_stability=structural_stability,
            activity_change=activity_change,
            lighting_change=lighting_change,
            rgi_delta=rgi_delta
        )
    
    def _calculate_time_delta(self, obs1: EvidencePackage, obs2: EvidencePackage) -> float:
        """Calculate time difference in hours"""
        # Simplified: assume observations are close in time
        # In production, parse timestamps
        return 0.5  # 30 minutes
    
    def _compare_objects(
        self,
        objects1: List[VisualObject],
        objects2: List[VisualObject]
    ) -> Tuple[List[Dict], List[Dict], List[Dict]]:
        """Compare visual objects between observations"""
        new_objects = []
        removed_objects = []
        changed_objects = []
        
        # Find new objects (in obs2 but not obs1)
        labels1 = {obj.label for obj in objects1}
        labels2 = {obj.label for obj in objects2}
        
        for obj in objects2:
            if obj.label not in labels1:
                new_objects.append({
                    "label": obj.label,
                    "confidence": obj.confidence,
                    "bbox": obj.bbox
                })
        
        # Find removed objects (in obs1 but not obs2)
        for obj in objects1:
            if obj.label not in labels2:
                removed_objects.append({
                    "label": obj.label,
                    "confidence": obj.confidence,
                    "bbox": obj.bbox
                })
        
        # Find changed objects (same label, different position/confidence)
        for obj1 in objects1:
            for obj2 in objects2:
                if obj1.label == obj2.label:
                    confidence_change = abs(obj1.confidence - obj2.confidence)
                    if confidence_change > 0.1:
                        changed_objects.append({
                            "label": obj1.label,
                            "confidence_change": confidence_change,
                            "old_confidence": obj1.confidence,
                            "new_confidence": obj2.confidence
                        })
        
        return new_objects, removed_objects, changed_objects
    
    def _calculate_structural_stability(
        self,
        obs1: EvidencePackage,
        obs2: EvidencePackage,
        new_objects: List[Dict],
        removed_objects: List[Dict]
    ) -> float:
        """Calculate structural stability (0-1)"""
        # Structural objects: buildings, street lights, roads, etc.
        structural_labels = {"building", "street_light", "road", "sidewalk", "bridge"}
        
        structural1 = {obj.label for obj in obs1.visual_objects if obj.label in structural_labels}
        structural2 = {obj.label for obj in obs2.visual_objects if obj.label in structural_labels}
        
        if not structural1:
            return 1.0
        
        # Calculate intersection
        common = structural1 & structural2
        stability = len(common) / len(structural1)
        
        return stability
    
    def _calculate_activity_change(self, obs1: EvidencePackage, obs2: EvidencePackage) -> float:
        """Calculate activity change (0-1)"""
        # Activity objects: cars, people, motorcycles, etc.
        activity_labels = {"car", "person", "motorcycle", "tuk-tuk", "bicycle"}
        
        activity1 = len([obj for obj in obs1.visual_objects if obj.label in activity_labels])
        activity2 = len([obj for obj in obs2.visual_objects if obj.label in activity_labels])
        
        if activity1 == 0 and activity2 == 0:
            return 0.0
        
        max_activity = max(activity1, activity2)
        change = abs(activity1 - activity2) / max_activity
        
        return change
    
    def _calculate_lighting_change(self, obs1: EvidencePackage, obs2: EvidencePackage) -> float:
        """Calculate lighting change (0-1)"""
        # Compare environmental signals
        lighting1 = None
        lighting2 = None
        
        for signal in obs1.environmental_signals:
            if signal.signal_type == "lighting":
                lighting1 = signal.value
        
        for signal in obs2.environmental_signals:
            if signal.signal_type == "lighting":
                lighting2 = signal.value
        
        if not lighting1 or not lighting2:
            return 0.0
        
        # Check if day/night changed
        is_night1 = lighting1.get("is_night", False)
        is_night2 = lighting2.get("is_night", False)
        
        if is_night1 != is_night2:
            return 1.0  # Day/night change is maximum change
        
        return 0.0
    
    def _calculate_rgi_delta(self, obs1: EvidencePackage, obs2: EvidencePackage) -> float:
        """Calculate change in Reality Gap Index"""
        # Simplified: compare number of defects/anomalies
        # In production, calculate full RGI for both
        
        defects1 = len([obj for obj in obs1.visual_objects if obj.confidence < 0.5])
        defects2 = len([obj for obj in obs2.visual_objects if obj.confidence < 0.5])
        
        return defects2 - defects1
    
    def analyze_location_history(
        self,
        location: Tuple[float, float],
        radius: float = 50.0
    ) -> Dict:
        """Analyze all observations at a location"""
        # Find observations near location
        nearby_observations = []
        for obs_id, evidence in self.observations.items():
            if evidence.metadata.gps_lat and evidence.metadata.gps_lng:
                distance = self._haversine_distance(
                    location[0], location[1],
                    evidence.metadata.gps_lat, evidence.metadata.gps_lng
                )
                if distance <= radius:
                    nearby_observations.append((obs_id, evidence))
        
        if len(nearby_observations) < 2:
            return {
                "status": "insufficient_data",
                "message": f"Only {len(nearby_observations)} observations at this location"
            }
        
        # Sort by timestamp
        nearby_observations.sort(key=lambda x: x[1].timestamp)
        
        # Compare consecutive observations
        comparisons = []
        for i in range(len(nearby_observations) - 1):
            obs_id_1 = nearby_observations[i][0]
            obs_id_2 = nearby_observations[i + 1][0]
            
            comparison = self.compare_observations(obs_id_1, obs_id_2)
            comparisons.append(comparison)
        
        # Calculate trends
        avg_stability = np.mean([c.structural_stability for c in comparisons])
        avg_activity_change = np.mean([c.activity_change for c in comparisons])
        avg_rgi_delta = np.mean([c.rgi_delta for c in comparisons])
        
        return {
            "status": "success",
            "observation_count": len(nearby_observations),
            "comparisons": len(comparisons),
            "trends": {
                "structural_stability": avg_stability,
                "activity_variability": avg_activity_change,
                "rgi_trend": avg_rgi_delta
            },
            "latest_observation": nearby_observations[-1][0],
            "recommendations": self._generate_recommendations(
                avg_stability, avg_activity_change, avg_rgi_delta
            )
        }
    
    def _generate_recommendations(
        self,
        stability: float,
        activity_change: float,
        rgi_delta: float
    ) -> List[str]:
        """Generate recommendations based on trends"""
        recommendations = []
        
        if stability < 0.8:
            recommendations.append("Structural changes detected - inspect infrastructure")
        
        if activity_change > 0.5:
            recommendations.append("High activity variability - monitor during different times")
        
        if rgi_delta > 0:
            recommendations.append("Reality gap increasing - maintenance needed")
        elif rgi_delta < 0:
            recommendations.append("Reality gap decreasing - improvements working")
        
        return recommendations
    
    def _haversine_distance(self, lat1, lng1, lat2, lng2) -> float:
        """Calculate distance between coordinates"""
        import math
        
        R = 6371000
        phi1 = math.radians(lat1)
        phi2 = math.radians(lat2)
        delta_phi = math.radians(lat2 - lat1)
        delta_lambda = math.radians(lng2 - lng1)
        
        a = math.sin(delta_phi / 2) ** 2 + \
            math.cos(phi1) * math.cos(phi2) * math.sin(delta_lambda / 2) ** 2
        c = 2 * math.atan2(math.sqrt(a), math.sqrt(1 - a))
        
        return R * c


# Example usage with Bangkok images
def analyze_bangkok_images():
    """Analyze all three Bangkok images"""
    print("=" * 60)
    print("BANGKOK TEMPORAL ANALYSIS")
    print("=" * 60)
    
    analyzer = TemporalAnalyzer()
    
    # Create evidence for three Bangkok images
    # Image 1: Tuk-tuk, street view
    evidence1 = EvidencePackage(
        image_id="bangkok_001",
        timestamp=datetime(2026, 6, 26, 20, 0, 0),
        metadata=ImageMetadata(gps_lat=13.7563, gps_lng=100.5018),
        visual_objects=[
            VisualObject("tuk-tuk", 0.85, [100, 300, 200, 400]),
            VisualObject("street_light", 0.80, [50, 50, 100, 400]),
            VisualObject("building", 0.90, [200, 100, 500, 400])
        ],
        semantic_objects=[],
        text_detections=[],
        geometric_features=[],
        environmental_signals=[
            EnvironmentalSignal("lighting", {"is_night": True}, 0.95)
        ],
        temporal_signals={}
    )
    
    # Image 2: Truck, motorcycles
    evidence2 = EvidencePackage(
        image_id="bangkok_002",
        timestamp=datetime(2026, 6, 26, 20, 15, 0),
        metadata=ImageMetadata(gps_lat=13.7565, gps_lng=100.5020),
        visual_objects=[
            VisualObject("truck", 0.88, [50, 250, 250, 450]),
            VisualObject("motorcycle", 0.82, [300, 350, 350, 400]),
            VisualObject("street_light", 0.85, [50, 50, 100, 400]),
            VisualObject("building", 0.92, [200, 100, 500, 400])
        ],
        semantic_objects=[],
        text_detections=[],
        geometric_features=[],
        environmental_signals=[
            EnvironmentalSignal("lighting", {"is_night": True}, 0.95)
        ],
        temporal_signals={}
    )
    
    # Image 3: Car, restaurant
    evidence3 = EvidencePackage(
        image_id="bangkok_003",
        timestamp=datetime(2026, 6, 26, 20, 30, 0),
        metadata=ImageMetadata(gps_lat=13.7564, gps_lng=100.5019),
        visual_objects=[
            VisualObject("car", 0.78, [100, 300, 200, 400]),
            VisualObject("street_light", 0.85, [50, 50, 100, 400]),
            VisualObject("building", 0.92, [200, 100, 500, 400]),
            VisualObject("sign", 0.88, [300, 50, 450, 150])
        ],
        semantic_objects=[],
        text_detections=[
            TextDetection("Sukhumvit Road", 0.92, [50, 50, 250, 100]),
            TextDetection("Sainokuni", 0.88, [300, 60, 450, 120])
        ],
        geometric_features=[],
        environmental_signals=[
            EnvironmentalSignal("lighting", {"is_night": True}, 0.95)
        ],
        temporal_signals={}
    )
    
    # Add to analyzer
    analyzer.add_observation("bangkok_001", evidence1)
    analyzer.add_observation("bangkok_002", evidence2)
    analyzer.add_observation("bangkok_003", evidence3)
    
    # Compare images
    print("\n📊 COMPARISON: Image 1 vs Image 2")
    comp1 = analyzer.compare_observations("bangkok_001", "bangkok_002")
    print(f"   Time delta: {comp1.time_delta}h")
    print(f"   New objects: {len(comp1.new_objects)}")
    for obj in comp1.new_objects:
        print(f"     - {obj['label']}")
    print(f"   Removed objects: {len(comp1.removed_objects)}")
    for obj in comp1.removed_objects:
        print(f"     - {obj['label']}")
    print(f"   Structural stability: {comp1.structural_stability:.1%}")
    print(f"   Activity change: {comp1.activity_change:.1%}")
    
    print("\n📊 COMPARISON: Image 2 vs Image 3")
    comp2 = analyzer.compare_observations("bangkok_002", "bangkok_003")
    print(f"   Time delta: {comp2.time_delta}h")
    print(f"   New objects: {len(comp2.new_objects)}")
    for obj in comp2.new_objects:
        print(f"     - {obj['label']}")
    print(f"   Removed objects: {len(comp2.removed_objects)}")
    for obj in comp2.removed_objects:
        print(f"     - {obj['label']}")
    print(f"   Structural stability: {comp2.structural_stability:.1%}")
    print(f"   Activity change: {comp2.activity_change:.1%}")
    
    # Analyze location history
    print("\n📈 LOCATION HISTORY ANALYSIS")
    history = analyzer.analyze_location_history((13.7564, 100.5019), radius=100)
    
    print(f"   Observations: {history['observation_count']}")
    print(f"   Comparisons: {history['comparisons']}")
    print(f"   Structural stability: {history['trends']['structural_stability']:.1%}")
    print(f"   Activity variability: {history['trends']['activity_variability']:.1%}")
    print(f"   RGI trend: {history['trends']['rgi_trend']:+.1f}")
    
    print("\n🎯 RECOMMENDATIONS")
    for rec in history['recommendations']:
        print(f"   - {rec}")
    
    return history


if __name__ == '__main__':
    analyze_bangkok_images()