songhaoyu/opi-yolo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

2026.8.20 视觉追踪器优化

Browse Source
This commit is contained in:
songhaoyu
2026-08-20 09:15:36 +08:00
parent 861a5981a6
commit ed0fc698f6
1 changed files with 53 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
drone_controller/target_tracker.py
+53 -14
View File
@@ -3,11 +3,12 @@ target_tracker.py — 轻量级多目标追踪过滤器
【解决的问题】
1. 目标短暂丢帧（2-5帧）导致无人机抽搐 → 卡尔曼滤波 coast/predict 平滑过渡
2. 多目标 ID 飘移导致甩头 → 中心距离 + IoU 加权关联锁定 ID 不变
2. 多目标 ID 飘移导致甩头 → 中心距离 + IoU 加权关联 + 主目标锁定
【核心设计】
- 恒定速度卡尔曼滤波器（Kalman Filter × 8状态：cx, cy, w, h, vx, vy, vw, vh）
- 最邻近贪婪关联（中心距离 + IoU 加权代价），提高交叉/靠近时的 ID 稳定性
- 主目标锁定：一旦选定 primary_id，只要该轨道还活着就持续锁定，避免被新高置信度目标抢走
- 死区预测（Coast/Predict）：连续丢失 ≤ max_lost_frames 帧时用滤波器预测值
- 超过 max_lost_frames → 释放轨道，对外判丢失
@@ -289,7 +290,8 @@ class TargetTracker:
职责：
- 接收每帧原始 YOLO 检测列表
- 通过卡尔曼滤波 + 最邻近关联维持稳定目标ID
- 通过卡尔曼滤波 + 中心距离/IoU 关联维持稳定目标ID
- 主目标一旦锁定就持续跟随，直到该轨道真正死亡才重新选择
- 提供平滑后的目标框 / 中心点供 PID 控制使用
- 自动回收超过丢失上限的旧轨道
@@ -343,6 +345,9 @@ class TargetTracker:
self._tracks = {} # {track_id: _Tracklet}
self._frame_count = 0
# 主目标锁定：一旦选定就持续锁定，直到该轨道死亡
self._locked_primary_id = None
# 缓存上一帧的 primary 结果（锁外创建副本供控制线程读取）
self._last_result = {
"tracked": False,
@@ -398,6 +403,7 @@ class TargetTracker:
matches = []
unmatched_tids = list(self._tracks.keys())
unmatched_dids = []
det_centers = []
# ---- 步骤 3：更新匹配的轨道 ----
for tid, did in matches:
@@ -412,6 +418,9 @@ class TargetTracker:
if self._tracks[tid].is_dead:
logger.debug("轨道 %d 已死亡 (lost=%d), 回收",
tid, self._tracks[tid].lost_count)
# 如果死的是锁定主目标，清除锁定
if tid == self._locked_primary_id:
self._locked_primary_id = None
del self._tracks[tid]
# ---- 步骤 5：未匹配的检测 → 创建新轨道 ----
@@ -430,7 +439,7 @@ class TargetTracker:
self._tracks[self._next_id] = new_track
self._next_id += 1
# ---- 步骤 6：选取主目标（优先级：置信度 > 命中数 > 贴近画面中心） ----
# ---- 步骤 6：选取主目标（带锁定逻辑） ----
result = self._select_primary(frame_shape)
self._last_result = result
return result
@@ -444,6 +453,7 @@ class TargetTracker:
self._tracks.clear()
self._next_id = 0
self._frame_count = 0
self._locked_primary_id = None
self._last_result = {
"tracked": False,
"primary_id": None,
@@ -587,9 +597,13 @@ class TargetTracker:
def _select_primary(self, frame_shape=None):
"""
从活跃轨道中选出主目标
从活跃轨道中选出主目标（带锁定）
优先级：
锁定逻辑：
1. 如果已有 _locked_primary_id，且该轨道仍存在 → 直接继续用它
2. 只有锁定目标真正消失（被回收）后，才重新按优先级选新主目标并锁定
新选主目标时的优先级：
1. 含有实测（非 coasting）且确认（hits ≥ min_hits）
2. 实测轨道中置信度最高
3. 若无实测轨道，取 coast 中最接近画面中心的
@@ -598,6 +612,7 @@ class TargetTracker:
dict — 与 update() 返回值格式一致
"""
if not self._tracks:
self._locked_primary_id = None
return {
"tracked": False,
"primary_id": None,
@@ -610,28 +625,47 @@ class TargetTracker:
"raw": None,
}
# 分组：实测轨道 vs coasting 轨道
# ---------- 优先使用已锁定的主目标 ----------
if self._locked_primary_id is not None and self._locked_primary_id in self._tracks:
track = self._tracks[self._locked_primary_id]
cx, cy = track.last_center
box = track.last_box
conf = track.confidence
lost = track.lost_count
is_predicted = track.lost_count > 0
return {
"tracked": True,
"primary_id": self._locked_primary_id,
"center": (float(cx), float(cy)),
"box": box,
"confidence": float(conf),
"is_predicted": is_predicted,
"lost_frames": lost,
"n_active": len(self._tracks),
"raw": (None if is_predicted else box),
}
# ---------- 没有锁定目标，或锁定目标已死 → 重新选择 ----------
measured = []
coasting = []
for tid, track in self._tracks.items():
entry = (tid, track)
if track.lost_count == 0 and track.hits >= self.min_hits:
measured.append(entry)
elif track.lost_count == 0:
# hits < min_hits 尚未确认，视为暂态
if track.lost_count == 0:
measured.append(entry)
else:
coasting.append(entry)
# 优先从实测轨道选
chosen = None
is_predicted = False
if measured:
# 置信度最高的实测轨道
measured.sort(key=lambda kv: kv[1].confidence, reverse=True)
chosen = measured[0]
# 优先选已确认（hits 足够）且置信度最高的
confirmed = [e for e in measured if e[1].hits >= self.min_hits]
pool = confirmed if confirmed else measured
pool.sort(key=lambda kv: kv[1].confidence, reverse=True)
chosen = pool[0]
elif coasting:
# 无实测时选最靠近画面中心的
cx_frame = frame_shape[1] / 2.0 if frame_shape else 320.0
@@ -644,6 +678,7 @@ class TargetTracker:
is_predicted = True
if chosen is None:
self._locked_primary_id = None
return {
"tracked": False,
"primary_id": None,
@@ -657,6 +692,9 @@ class TargetTracker:
}
tid, track = chosen
# 锁定新主目标
self._locked_primary_id = tid
cx, cy = track.last_center
box = track.last_box
conf = track.confidence
@@ -678,4 +716,5 @@ class TargetTracker:
tracks_info = ", ".join(str(t) for t in self._tracks.values())
return (f"TargetTracker(frame={self._frame_count}, "
f"active={len(self._tracks)}, "
f"locked={self._locked_primary_id}, "
f"tracks=[{tracks_info}])")