File size: 9,941 Bytes
7b18843
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
#!/usr/bin/env python3
"""
preprocess_step1_correct_ply.py
================================
NEST3D Pre-processing Step 1: PLY correction.

For each sample in <data-dir>/sampleXXX/sampleXXX.ply:
  Fix 1 - Outlier grass removal:
      Keep all tree+nest points.
      Keep only grass within 5m XY of any tree/nest point.
  Fix 2 - Ground plane leveling (samples 015, 061, 063 only):
      RANSAC plane fit on grass -> rotate to horizontal.
  Fix 3 - Z grounding (all samples):
      Subtract 5th percentile of grass Z -> ground becomes ~0.
  Fix 4 - Point budget (all samples):
      If total > 10M: keep ALL nest, keep 1M grass min, downsample tree.

Points labeled 255 (unclassified/ignore) are preserved unchanged throughout
all four fixes -- never removed, downsampled, or reassigned.

Input:  <data-dir>/sampleXXX/sampleXXX.ply
Output: <data-dir>/sampleXXX/sampleXXX_corrected.ply

Labels (CloudCompare):
  0   = grass
  1   = tree
  2   = nest
  255 = unclassified / ignore

Usage:
  python preprocess_step1_correct_ply.py --data-dir /path/to/reconstructions

Author: NEST3D team
"""

import argparse
import numpy as np
import json
from pathlib import Path
from plyfile import PlyData, PlyElement
from sklearn.neighbors import BallTree

# ── Config ────────────────────────────────────────────────────────────────────
GRASS_KEEP_RADIUS_M  = 5.0
MAX_TOTAL_POINTS     = 10_000_000
GRASS_MINIMUM        = 1_000_000
DOWNSAMPLE_SEED      = 42
Z_GROUND_PERCENTILE  = 5

NEST_LABEL = 2
TREE_LABEL = 1

TILT_SAMPLES = {
    "sample015": {"ransac_min_inlier_frac": 0.5},
    "sample061": {"ransac_min_inlier_frac": 0.5},
    "sample063": {"ransac_min_inlier_frac": 0.15},
}
RANSAC_ITERATIONS  = 1000
RANSAC_THRESHOLD_M = 0.08

# ── I/O ───────────────────────────────────────────────────────────────────────
def load_ply(path):
    ply = PlyData.read(str(path))
    v = ply["vertex"]
    xyz      = np.stack([v["x"], v["y"], v["z"]], axis=1).astype(np.float64)
    rgb      = np.stack([v["red"], v["green"], v["blue"]], axis=1)
    lbl      = np.array(v["scalar_Classification"], dtype=np.int32)
    orig_idx = np.array(v["scalar_Original_cloud_index"], dtype=np.int32)
    return xyz, rgb, lbl, orig_idx

def save_ply(path, xyz, rgb, lbl, orig_idx):
    n = len(xyz)
    arr = np.zeros(n, dtype=[
        ("x","f4"),("y","f4"),("z","f4"),
        ("red","u1"),("green","u1"),("blue","u1"),
        ("scalar_Classification","f4"),
        ("scalar_Original_cloud_index","f4"),
    ])
    arr["x"], arr["y"], arr["z"] = xyz[:,0], xyz[:,1], xyz[:,2]
    arr["red"], arr["green"], arr["blue"] = rgb[:,0], rgb[:,1], rgb[:,2]
    arr["scalar_Classification"]       = lbl.astype(np.float32)
    arr["scalar_Original_cloud_index"] = orig_idx.astype(np.float32)
    PlyData([PlyElement.describe(arr,"vertex")], text=False).write(str(path))

# ── Fix 1: Outlier grass removal ──────────────────────────────────────────────
def remove_isolated_grass(xyz, rgb, lbl, orig_idx):
    tree_nest = (lbl==TREE_LABEL)|(lbl==NEST_LABEL)
    grass     = (lbl==0)
    other     = ~(grass|tree_nest)

    if tree_nest.sum()==0 or grass.sum()==0:
        return xyz, rgb, lbl, orig_idx, 0

    bt = BallTree(xyz[tree_nest,:2], metric="euclidean")
    dist,_ = bt.query(xyz[grass,:2], k=1)
    near = dist[:,0] <= GRASS_KEEP_RADIUS_M

    keep = np.zeros(len(xyz), bool)
    keep[tree_nest] = True
    keep[other]     = True
    gi = np.where(grass)[0]
    keep[gi[near]]  = True
    removed = int((~keep).sum())
    return xyz[keep], rgb[keep], lbl[keep], orig_idx[keep], removed

# ── Fix 2: Ground leveling ────────────────────────────────────────────────────
def fit_plane_ransac(pts, min_inlier_frac):
    best_normal, best_d, best_n = None, None, 0
    min_inliers = int(min_inlier_frac * len(pts))
    rng = np.random.default_rng(DOWNSAMPLE_SEED)
    for _ in range(RANSAC_ITERATIONS):
        idx = rng.choice(len(pts), 3, replace=False)
        p0,p1,p2 = pts[idx]
        normal = np.cross(p1-p0, p2-p0)
        nl = np.linalg.norm(normal)
        if nl < 1e-8: continue
        normal /= nl
        d = -normal @ p0
        if normal[2] < 0: normal,d = -normal,-d
        n_in = int((np.abs(pts@normal+d) < RANSAC_THRESHOLD_M).sum())
        if n_in > best_n:
            best_n, best_normal, best_d = n_in, normal.copy(), d
    return best_normal, best_d, best_n >= min_inliers

def rotation_to_up(normal):
    up = np.array([0.,0.,1.])
    normal = normal/np.linalg.norm(normal)
    axis = np.cross(normal, up)
    al = np.linalg.norm(axis)
    if al < 1e-8:
        return np.eye(3) if normal[2]>0 else np.diag([1.,-1.,-1.])
    axis /= al
    angle = np.arccos(np.clip(np.dot(normal,up),-1,1))
    K = np.array([[0,-axis[2],axis[1]],[axis[2],0,-axis[0]],[-axis[1],axis[0],0]])
    return np.eye(3)+np.sin(angle)*K+(1-np.cos(angle))*(K@K)

def level_ground(xyz, lbl, min_inlier_frac):
    grass_pts = xyz[lbl==0]
    if len(grass_pts) < 100:
        return xyz, False
    normal, d, ok = fit_plane_ransac(grass_pts, min_inlier_frac)
    if not ok:
        return xyz, False
    R = rotation_to_up(normal)
    return (R @ xyz.T).T, True

# ── Fix 3: Z grounding ────────────────────────────────────────────────────────
def ground_z(xyz, lbl):
    grass_z = xyz[lbl==0, 2]
    z_ref = np.percentile(grass_z if len(grass_z)>0 else xyz[:,2], Z_GROUND_PERCENTILE)
    xyz = xyz.copy()
    xyz[:,2] -= z_ref
    return xyz, float(z_ref)

# ── Fix 4: Point budget ───────────────────────────────────────────────────────
def apply_budget(xyz, rgb, lbl, orig_idx):
    if len(xyz) <= MAX_TOTAL_POINTS:
        return xyz, rgb, lbl, orig_idx, 0, 0

    nest_mask  = (lbl==NEST_LABEL)
    tree_mask  = (lbl==TREE_LABEL)
    grass_mask = (lbl==0)
    other_mask = ~(nest_mask|tree_mask|grass_mask)

    n_nest  = int(nest_mask.sum())
    n_tree  = int(tree_mask.sum())
    n_grass = int(grass_mask.sum())
    n_other = int(other_mask.sum())

    n_grass_keep = min(GRASS_MINIMUM, n_grass)
    n_tree_keep  = MAX_TOTAL_POINTS - n_nest - n_grass_keep - n_other

    rng = np.random.default_rng(DOWNSAMPLE_SEED)
    keep = np.zeros(len(xyz), bool)
    keep[nest_mask]  = True
    keep[other_mask] = True

    gi = np.where(grass_mask)[0]
    keep[rng.choice(gi, n_grass_keep, replace=False)] = True

    ti = np.where(tree_mask)[0]
    if n_tree_keep > 0:
        n_tree_keep = min(n_tree_keep, n_tree)
        keep[rng.choice(ti, n_tree_keep, replace=False)] = True

    grass_rm = n_grass - n_grass_keep
    tree_rm  = max(0, n_tree - n_tree_keep)
    return xyz[keep], rgb[keep], lbl[keep], orig_idx[keep], grass_rm, tree_rm

# ── Main ──────────────────────────────────────────────────────────────────────
def main():
    parser = argparse.ArgumentParser(description="NEST3D Step 1: PLY correction")
    parser.add_argument(
        "--data-dir", type=Path, default=Path("./reconstructions"),
        help="Path to the reconstructions/ folder containing sampleXXX subfolders (default: ./reconstructions)"
    )
    args = parser.parse_args()
    recon_dir = args.data_dir

    sample_dirs = sorted(recon_dir.glob("sample*"))
    print(f"Found {len(sample_dirs)} samples in {recon_dir}\n")

    log = {}
    for sample_dir in sample_dirs:
        sample_id = sample_dir.name
        in_ply    = sample_dir / f"{sample_id}.ply"
        out_ply   = sample_dir / f"{sample_id}_corrected.ply"

        if not in_ply.exists():
            print(f"[SKIP] {sample_id}: no PLY found")
            continue
        if out_ply.exists():
            print(f"[SKIP] {sample_id}: already corrected")
            continue

        print(f"[PROC] {sample_id}")
        xyz, rgb, lbl, orig_idx = load_ply(in_ply)
        n_before = len(xyz)

        # Fix 1
        print(f"  Fix1 grass outliers ...", end=" ")
        xyz, rgb, lbl, orig_idx, grass_rm = remove_isolated_grass(xyz, rgb, lbl, orig_idx)
        print(f"removed {grass_rm:,}")

        # Fix 2
        if sample_id in TILT_SAMPLES:
            print(f"  Fix2 leveling ...", end=" ")
            cfg = TILT_SAMPLES[sample_id]
            xyz, ok = level_ground(xyz, lbl, cfg["ransac_min_inlier_frac"])
            print("done" if ok else "FAILED")

        # Fix 3
        print(f"  Fix3 Z grounding ...", end=" ")
        xyz, z_ref = ground_z(xyz, lbl)
        print(f"z_ref={z_ref:.3f}m")

        # Fix 4
        print(f"  Fix4 budget ({len(xyz):,} pts) ...", end=" ")
        xyz, rgb, lbl, orig_idx, g_rm, t_rm = apply_budget(xyz, rgb, lbl, orig_idx)
        print(f"grass-{g_rm:,} tree-{t_rm:,} -> {len(xyz):,}")

        save_ply(out_ply, xyz, rgb, lbl, orig_idx)

        log[sample_id] = {
            "n_before": n_before, "n_after": len(xyz),
            "grass_outliers_removed": grass_rm,
            "grass_downsampled": g_rm, "tree_downsampled": t_rm,
            "z_ref": z_ref,
        }
        print(f"  -> saved {out_ply.name}\n")

    log_path = recon_dir / "correction_log.json"
    with open(log_path, "w") as f:
        json.dump(log, f, indent=2)
    print(f"Done! Log saved to {log_path}")

if __name__ == "__main__":
    main()