验收1

fish_api/app/services/measure.py

@@ -11,6 +11,8 @@ from datetime import datetime, timezone
 from pathlib import Path
 from typing import Any, Dict, List, Optional, Tuple
 
+import numpy as np
+
 from app.logging_config import format_json_pretty
 from app.settings import Settings
 from app.state import MeasureSnapshot
@@ -187,12 +189,15 @@ def _safe_media_prefix(stem: str) -> str:
 
 
 def _result_from_weight_prediction(data: Dict[str, Any]) -> List[Dict[str, Any]]:
-    """按 track_id 聚合：体重取 max(predicted_weight_g)，体长取达到 max 的那条 PLY 的 length_input (mm)。"""
+    """按 track_id 聚合：对每个 track_id 的点云，取 predicted_weight_g 最大的 top5 平均值作为体重，
+    体长取这 top5 点云的平均 length_input (mm)。只使用通过点云分类器的好点云（used_for_prediction=True）。"""
     items = data.get("per_cloud") or data.get("per_file") or []
     if not isinstance(items, list):
         return []
-    # tid -> (max_weight_g, length_mm at max weight)
-    best: Dict[int, Tuple[float, float]] = {}
+
+    # tid -> list of (weight_g, length_mm) for this track
+    track_predictions: Dict[int, List[Tuple[float, float]]] = {}
+
     for it in items:
         if not isinstance(it, dict):
             continue
@@ -202,24 +207,55 @@ def _result_from_weight_prediction(data: Dict[str, Any]) -> List[Dict[str, Any]]
         tid = _parse_tid_from_ply_name(Path(str(ply)).name)
         if tid is None:
             continue
+
+        # 只使用通过点云分类器的好点云
+        if not it.get("used_for_prediction", True):
+            continue
+
         try:
             wg = float(it.get("predicted_weight_g", float("nan")))
         except (TypeError, ValueError):
             continue
         if not math.isfinite(wg):
             continue
+
         try:
             ln = float(it.get("length_input", float("nan")))
         except (TypeError, ValueError):
             ln = float("nan")
-        if tid not in best or wg > best[tid][0]:
-            best[tid] = (wg, ln)
+
+        if tid not in track_predictions:
+            track_predictions[tid] = []
+        track_predictions[tid].append((wg, ln))
+
     out: List[Dict[str, Any]] = []
-    for tid in sorted(best.keys()):
-        wg, ln = best[tid]
-        if not math.isfinite(ln):
+    TOP_K = 5
+
+    for tid in sorted(track_predictions.keys()):
+        predictions = track_predictions[tid]
+        if not predictions:
             continue
-        out.append({"id": tid, "weight": wg, "length": ln})
+
+        # 按重量降序排序，取 top5
+        predictions_sorted = sorted(predictions, key=lambda x: x[0], reverse=True)
+        top5 = predictions_sorted[:min(TOP_K, len(predictions_sorted))]
+
+        # 计算 top5 平均重量和平均长度
+        avg_weight = float(np.mean([p[0] for p in top5]))
+        valid_lengths = [p[1] for p in top5 if math.isfinite(p[1])]
+        avg_length = float(np.mean(valid_lengths)) if valid_lengths else float("nan")
+
+        if not math.isfinite(avg_length):
+            continue
+
+        out.append({
+            "id": tid,
+            "type": "大黄鱼",
+            "weight": str(round(avg_weight)),
+            "length": str(round(avg_length)),
+            "date": datetime.now(timezone.utc).strftime("%Y-%m-%d"),
+        })
+
     return out