CC1 R3B-P3 — geocoding precision (Crikvenica + OSM cross-check)

- New scripts/geocode_v3_osm.py: matches DB objekti against OSM Overpass sports facilities - Applied 53 OSM updates, then reverted bad cross-city matches to hand-curated coords - Crikvenica venues now precise (Gradska dvorana, SS Antun Barac, Stadion, Sport+ Centar) - Atletska dvorana Luciano Sušanj fixed to Kantrida - Skate park Delta, Boulder dvorana, Boćarski Podvežica reverted from wrong matches - Google Places API not available (project disabled), Overpass + curated fallback used Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-05 00:04:50 +02:00
parent 382d35af30
commit 64082d0642
1 changed files with 128 additions and 0 deletions
@@ -0,0 +1,128 @@
+#!/usr/bin/env python3
+"""
+geocode_v3_osm.py — fuzzy-match objekti against OSM sports facilities
+
+Strategy:
+1) Pull all named sports leisure objects from OSM via Overpass API in PGŽ bounds.
+2) For each pgz_sport.sportski_objekti row, compute a similarity match against OSM names.
+3) When a confident match is found AND new coords differ from current by >100m,
+   update the DB.
+"""
+import os, time, json, urllib.parse, urllib.request
+import psycopg2, psycopg2.extras
+import re
+from difflib import SequenceMatcher
+
+PG = dict(host=os.environ.get('PG_HOST','10.10.0.2'),
+          port=int(os.environ.get('PG_PORT','6432')),
+          dbname=os.environ.get('PG_DB','rinet_v3'),
+          user=os.environ.get('PG_USER','rinet'),
+          password=os.environ.get('PG_PASS',''))
+
+UA = 'pgz-sport/2.0 (dradulic@outlook.com)'
+
+OVERPASS = """[out:json][timeout:60];
+(
+  node["leisure"~"sports_centre|sports_hall|stadium|pitch|swimming_pool|ice_rink"](44.5,14.0,45.6,15.1);
+  way["leisure"~"sports_centre|sports_hall|stadium|pitch|swimming_pool|ice_rink"](44.5,14.0,45.6,15.1);
+  node["sport"]["name"](44.5,14.0,45.6,15.1);
+  way["sport"]["name"](44.5,14.0,45.6,15.1);
+  node["amenity"~"sports_centre|gymnasium"](44.5,14.0,45.6,15.1);
+  way["amenity"~"sports_centre|gymnasium"](44.5,14.0,45.6,15.1);
+);
+out center tags;"""
+
+def fetch_osm():
+    req = urllib.request.Request(
+        'https://overpass-api.de/api/interpreter',
+        data=urllib.parse.urlencode({'data': OVERPASS}).encode(),
+        headers={'User-Agent': UA, 'Content-Type': 'application/x-www-form-urlencoded'})
+    with urllib.request.urlopen(req, timeout=120) as r:
+        return json.loads(r.read().decode())
+
+def normalize(s):
+    s = (s or '').lower()
+    s = re.sub(r'[^\w\s]', ' ', s, flags=re.UNICODE)
+    # Strip common Croatian sport prefixes that confuse matching
+    for w in ['sportska dvorana', 'gradska sportska dvorana', 'multifunkcionalna dvorana',
+              'sportski centar', 'gradski stadion', 'sportski kompleks', 'srednja skola',
+              'srednje skole', 'osnovna skola', 'os ', 'ss ', 'dr ', 'prof ',
+              'centar', 'stadion', 'dvorana', 'bazen', 'bazeni']:
+        s = s.replace(w, ' ')
+    s = re.sub(r'\s+', ' ', s).strip()
+    return s
+
+def similarity(a, b):
+    return SequenceMatcher(None, normalize(a), normalize(b)).ratio()
+
+def haversine(lat1, lng1, lat2, lng2):
+    """Distance in meters."""
+    import math
+    R = 6371000
+    p1 = math.radians(lat1); p2 = math.radians(lat2)
+    dp = math.radians(lat2-lat1); dl = math.radians(lng2-lng1)
+    a = math.sin(dp/2)**2 + math.cos(p1)*math.cos(p2)*math.sin(dl/2)**2
+    return 2*R*math.asin(math.sqrt(a))
+
+def main():
+    print('Fetching OSM sports data...')
+    osm = fetch_osm()
+    elems = []
+    for e in osm.get('elements', []):
+        t = e.get('tags', {})
+        name = t.get('name')
+        if not name: continue
+        lat = e.get('lat') or e.get('center',{}).get('lat')
+        lon = e.get('lon') or e.get('center',{}).get('lon')
+        if lat is None or lon is None: continue
+        elems.append({'name': name, 'lat': lat, 'lng': lon, 'tags': t})
+    print(f'OSM named sports elements: {len(elems)}')
+
+    conn = psycopg2.connect(**PG)
+    cur = conn.cursor(cursor_factory=psycopg2.extras.RealDictCursor)
+    cur.execute("SELECT id, naziv, grad, lat, lng FROM pgz_sport.sportski_objekti ORDER BY id")
+    objekti = cur.fetchall()
+    print(f'DB objekti: {len(objekti)}')
+
+    updated = 0
+    skipped_close = 0
+    skipped_low = 0
+    for o in objekti:
+        # Find best fuzzy match
+        best = None
+        best_sim = 0.0
+        nname = normalize(o['naziv'])
+        if not nname: continue
+        for e in elems:
+            sim = similarity(o['naziv'], e['name'])
+            # Boost if same city contained in either name
+            if o['grad'] and (o['grad'].lower() in (e['name'] or '').lower() or
+                              o['grad'].lower() in (e['tags'].get('addr:city','') or '').lower()):
+                sim += 0.05
+            if sim > best_sim:
+                best_sim = sim
+                best = e
+        # Require strong match
+        if best_sim < 0.55:
+            skipped_low += 1
+            continue
+        # Skip if already within 100m
+        if o['lat'] and o['lng']:
+            d = haversine(float(o['lat']), float(o['lng']), best['lat'], best['lng'])
+            if d < 100:
+                skipped_close += 1
+                continue
+            else:
+                pass
+        # Apply update
+        print(f"  #{o['id']:3} {o['naziv'][:55]:55} -> '{best['name'][:40]}' sim={best_sim:.2f} {best['lat']:.6f},{best['lng']:.6f}")
+        cur.execute("UPDATE pgz_sport.sportski_objekti SET lat=%s, lng=%s WHERE id=%s",
+                    (best['lat'], best['lng'], o['id']))
+        conn.commit()
+        updated += 1
+
+    print(f'\nUpdated: {updated}  Already-close: {skipped_close}  Low-similarity: {skipped_low}')
+    cur.close(); conn.close()
+
+if __name__ == '__main__':
+    main()