260705_2

2026-07-05 21:27:23 +09:00
parent 23d907265a
commit 3abc2edba6
83 changed files with 10351 additions and 1217 deletions
@@ -0,0 +1,112 @@
+"""B04 LAS/LAZ 고속 구조화 엔진."""
+
+import os
+import tempfile
+from collections.abc import Callable
+from pathlib import Path
+
+import laspy
+import numpy as np
+
+from config.config_system import SURFACE_DEFAULT_RGB_VALUE, SURFACE_LAS_CHUNK_SIZE
+
+
+def structurize_las(
+    las_path: str | Path,
+    output_dir: str | Path,
+    progress_callback: Callable[[int], None] | None = None,
+) -> Path:
+    """LAS/LAZ 속성을 청크로 읽어 B04 structured.npz로 원자적 저장한다."""
+    source = Path(las_path)
+    target_dir = Path(output_dir)
+    target_dir.mkdir(parents=True, exist_ok=True)
+    target = target_dir / "structured.npz"
+
+    with laspy.open(source) as las_file:
+        header = las_file.header
+        total_points = int(header.point_count)
+        point_format = header.point_format
+        dimensions = set(point_format.dimension_names)
+        has_rgb = {"red", "green", "blue"}.issubset(dimensions)
+        has_intensity = "intensity" in dimensions
+        has_returns = {"return_number", "number_of_returns"}.issubset(dimensions)
+        has_classification = "classification" in dimensions
+        bounds = np.array(
+            [
+                [float(header.mins[0]), float(header.maxs[0])],
+                [float(header.mins[1]), float(header.maxs[1])],
+                [float(header.mins[2]), float(header.maxs[2])],
+            ],
+            dtype=np.float64,
+        )
+
+        xyz = np.empty((total_points, 3), dtype=np.float64)
+        intensity = np.zeros(total_points, dtype=np.uint16)
+        rgb = np.full((total_points, 3), SURFACE_DEFAULT_RGB_VALUE, dtype=np.uint8)
+        return_number = np.ones(total_points, dtype=np.uint8)
+        number_of_returns = np.ones(total_points, dtype=np.uint8)
+        classification = np.zeros(total_points, dtype=np.uint8)
+
+        offset = 0
+        for chunk in las_file.chunk_iterator(SURFACE_LAS_CHUNK_SIZE):
+            chunk_size = len(chunk)
+            section = slice(offset, offset + chunk_size)
+            xyz[section, 0] = np.asarray(chunk.x, dtype=np.float64)
+            xyz[section, 1] = np.asarray(chunk.y, dtype=np.float64)
+            xyz[section, 2] = np.asarray(chunk.z, dtype=np.float64)
+            if has_intensity:
+                intensity[section] = np.asarray(chunk.intensity, dtype=np.uint16)
+            if has_rgb:
+                colors = np.stack(
+                    [
+                        np.asarray(chunk.red, dtype=np.float64),
+                        np.asarray(chunk.green, dtype=np.float64),
+                        np.asarray(chunk.blue, dtype=np.float64),
+                    ],
+                    axis=1,
+                )
+                if colors.size and float(colors.max()) > 255.0:
+                    colors /= 256.0
+                rgb[section] = colors.clip(0, 255).astype(np.uint8)
+            if has_returns:
+                return_number[section] = np.asarray(chunk.return_number, dtype=np.uint8)
+                number_of_returns[section] = np.asarray(chunk.number_of_returns, dtype=np.uint8)
+            if has_classification:
+                classification[section] = np.asarray(chunk.classification, dtype=np.uint8)
+            offset += chunk_size
+            if progress_callback:
+                progress_callback(int(offset / total_points * 100) if total_points else 100)
+
+    temporary_path: Path | None = None
+    try:
+        with tempfile.NamedTemporaryFile(
+            mode="wb",
+            dir=target_dir,
+            prefix=".structured.",
+            suffix=".npz.tmp",
+            delete=False,
+        ) as temporary:
+            temporary_path = Path(temporary.name)
+            np.savez_compressed(
+                temporary,
+                xyz=xyz,
+                intensity=intensity,
+                rgb=rgb,
+                return_number=return_number,
+                number_of_returns=number_of_returns,
+                classification=classification,
+                bounds=bounds,
+                total_points=np.array([total_points], dtype=np.int64),
+                has_rgb=np.array([int(has_rgb)], dtype=np.int8),
+            )
+            temporary.flush()
+            os.fsync(temporary.fileno())
+        os.replace(temporary_path, target)
+        temporary_path = None
+    finally:
+        if temporary_path is not None:
+            temporary_path.unlink(missing_ok=True)
+
+    if progress_callback and total_points == 0:
+        progress_callback(100)
+    return target