File size: 7,345 Bytes
1b4b6d2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b6675ab
 
 
 
 
1b4b6d2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1c70885
 
 
1b4b6d2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
"""
استخراج النصوص والأرقام من المخططات باستخدام OCR
"""
try:
    import easyocr
except ImportError:
    easyocr = None
import numpy as np
import re
import cv2
from typing import List, Dict, Optional
from functools import lru_cache


@lru_cache(maxsize=1)
def get_ocr_reader():
    """
    يحمل نموذج OCR مرة واحدة فقط (cache)
    يدعم العربي والإنجليزي
    """
    try:
        import easyocr
        return easyocr.Reader(['en', 'ar'], gpu=False)
    except ImportError:
        return None


def _preprocess_for_ocr(img: np.ndarray) -> np.ndarray:
    """
    Denoise + local-contrast (CLAHE) so small/blurry dimension text on dense
    construction drawings becomes more legible. Keeps the SAME image dimensions,
    so the bbox coordinates OCR returns remain valid (no rescaling needed).
    """
    try:
        if img is None:
            return img
        g = img
        if len(g.shape) == 3:
            g = cv2.cvtColor(g, cv2.COLOR_BGR2GRAY)
        g = cv2.fastNlMeansDenoising(g, None, 7, 7, 21)
        g = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8)).apply(g)
        return g
    except Exception:
        return img


def extract_all_text(page_array: np.ndarray, preprocess: bool = True) -> List[Dict]:
    """
    يستخرج كل النص من الصورة
    Returns: list of {text, bbox, confidence}
    bbox format: [[x1,y1],[x2,y1],[x2,y2],[x1,y2]]
    """
    reader = get_ocr_reader()
    if reader is None:
        return []
        
    img = _preprocess_for_ocr(page_array) if preprocess else page_array
    results = reader.readtext(img)

    extracted = []
    for bbox, text, confidence in results:
        if confidence > 0.3:  # filter low confidence
            extracted.append({
                "text":       text.strip(),
                "bbox":       bbox,
                "confidence": round(confidence, 3),
                "center_x":   int((bbox[0][0] + bbox[2][0]) / 2),
                "center_y":   int((bbox[0][1] + bbox[2][1]) / 2),
            })

    return extracted


def extract_dimensions(text_results: List[Dict]) -> List[Dict]:
    """
    يفلتر النتائج ويستخرج فقط الأرقام التي تمثل أبعاداً
    يبحث عن: أرقام عشرية، أرقام بوحدة m أو cm أو mm
    Returns: list of {value_meters, text, bbox, center_x, center_y}
    """
    dimension_pattern = re.compile(
        r"""
        (\d+\.?\d*)     # number (integer or decimal)
        \s*             # optional space
        (mm|cm|m)?      # optional unit — longest match first
        """,
        re.VERBOSE | re.IGNORECASE,
    )

    dimensions = []

    for item in text_results:
        text = item["text"].strip()

        # Skip text that's clearly not a dimension
        if len(text) > 15:  # too long to be a dimension
            continue
        if re.search(r'[a-zA-Z]{3,}', text):  # has long words
            continue

        matches = dimension_pattern.findall(text)

        for num_str, unit in matches:
            try:
                value = float(num_str)

                # Convert to meters
                if unit.lower() == 'cm':
                    value_m = value / 100
                elif unit.lower() == 'mm':
                    value_m = value / 1000
                else:
                    # If no unit, check magnitude
                    if value > 100:   # likely mm
                        value_m = value / 1000
                    elif value > 20:  # likely cm
                        value_m = value / 100
                    else:             # likely meters
                        value_m = value

                # Filter unrealistic dimensions for villas (0.1m – 50m)
                if 0.1 <= value_m <= 50.0:
                    dimensions.append({
                        "value_meters":  round(value_m, 3),
                        "original_text": text,
                        "unit":          unit if unit else "m (assumed)",
                        "bbox":          item["bbox"],
                        "center_x":      item["center_x"],
                        "center_y":      item["center_y"],
                        "confidence":    item["confidence"],
                    })
            except ValueError:
                continue

    return dimensions


def extract_dimensions_from_plain_text(text: str) -> List[Dict]:
    """
    Extract dimensions directly from plain text (e.g. fitz-extracted PDF text).
    Returns same format as extract_dimensions() but without bbox/center fields.
    """
    dimension_pattern = re.compile(
        r'(\d+\.?\d*)\s*(mm|cm|m)?(?=\s|$|[,;\)])',
        re.IGNORECASE,
    )
    dims = []
    seen = set()
    for line in text.splitlines():
        line = line.strip()
        if not line or len(line) > 30:
            continue
        if re.search(r'[a-zA-Z]{4,}', line):
            continue
        for num_str, unit in dimension_pattern.findall(line):
            try:
                value = float(num_str)
                unit_l = unit.lower()
                if unit_l == 'cm':
                    value_m = value / 100
                elif unit_l == 'mm':
                    value_m = value / 1000
                else:
                    if value > 100:
                        value_m = value / 1000
                    elif value > 20:
                        value_m = value / 100
                    else:
                        value_m = value
                value_m = round(value_m, 3)
                if 0.1 <= value_m <= 50.0 and value_m not in seen:
                    seen.add(value_m)
                    dims.append({
                        "value_meters":  value_m,
                        "original_text": line,
                        "unit":          unit if unit else "m (assumed)",
                        "bbox":          None,
                        "center_x":      0,
                        "center_y":      0,
                        "confidence":    0.85,
                    })
            except ValueError:
                continue
    return dims


def detect_scale_text_ocr(text_results: List[Dict]) -> Optional[str]:
    """
    يبحث عن نص المقياس مثل 1:100 أو SCALE 1:200
    """
    # Accept 1:100, 1 / 100, 1-100, and "SCALE 1:100" (case-insensitive)
    scale_pattern = re.compile(r'1\s*[:/\-]\s*\d{1,4}')

    for item in text_results:
        text = item["text"]
        if scale_pattern.search(text):
            return text

    return None


def draw_dimensions_on_image(
    page_array: np.ndarray,
    dimensions: List[Dict],
) -> np.ndarray:
    """
    يرسم الأبعاد المكتشفة على الصورة للمراجعة البصرية
    """
    img = page_array.copy()

    for dim in dimensions:
        bbox = dim["bbox"]
        x1, y1 = int(bbox[0][0]), int(bbox[0][1])
        x2, y2 = int(bbox[2][0]), int(bbox[2][1])

        # Draw green rectangle around detected dimension
        cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2)

        # Add label
        label = f"{dim['value_meters']}m"
        cv2.putText(
            img, label,
            (x1, y1 - 5),
            cv2.FONT_HERSHEY_SIMPLEX,
            0.5, (0, 200, 0), 1,
        )

    return img