Create mouse_tracker.py
Browse files- mouse_tracker.py +572 -0
mouse_tracker.py
ADDED
@@ -0,0 +1,572 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import cv2
|
3 |
+
import numpy as np
|
4 |
+
import matplotlib.pyplot as plt
|
5 |
+
import pandas as pd
|
6 |
+
import collections
|
7 |
+
import tempfile
|
8 |
+
from ultralytics import YOLO
|
9 |
+
import math
|
10 |
+
|
11 |
+
class MouseTrackerAnalyzer:
|
12 |
+
"""基于Ultralytics对象跟踪的鼠强迫游泳实验挣扎度分析器"""
|
13 |
+
def __init__(self, model_path, history_size=5, conf=0.25, iou=0.45, max_det=20, verbose=False):
|
14 |
+
# 初始化模型和参数
|
15 |
+
self.model = YOLO(model_path, task="segment", verbose=False)
|
16 |
+
self.history_size = history_size
|
17 |
+
self.verbose = verbose # 控制日志输出级别
|
18 |
+
self.struggle_threshold = 0.3 # 挣扎阈值
|
19 |
+
|
20 |
+
# 跟踪相关参数
|
21 |
+
self.conf = conf # 置信度阈值
|
22 |
+
self.iou = iou # IOU阈值
|
23 |
+
self.max_det = max_det # 最大检测数量
|
24 |
+
|
25 |
+
# 预设16种固定颜色 (BGR顺序)
|
26 |
+
self.colors = [
|
27 |
+
(255, 0, 0), # 红
|
28 |
+
(0, 255, 0), # 绿
|
29 |
+
(0, 0, 255), # 蓝
|
30 |
+
(255, 255, 0), # 青
|
31 |
+
(255, 0, 255), # 洋红
|
32 |
+
(0, 255, 255), # 黄
|
33 |
+
(128, 0, 0), # 深红
|
34 |
+
(128, 0, 128), # 紫
|
35 |
+
(0, 128, 128), # 青绿
|
36 |
+
(192, 192, 192),# 银
|
37 |
+
(128, 128, 128),# 灰
|
38 |
+
(255, 128, 0), # 橙
|
39 |
+
(255, 0, 128), # 粉
|
40 |
+
(0, 128, 255), # 浅蓝
|
41 |
+
(128, 255, 0), # 黄绿
|
42 |
+
(0, 255, 128) # 浅绿
|
43 |
+
]
|
44 |
+
# 追踪相关
|
45 |
+
self.prev_masks = {} # 上一帧各 ID 二值掩码
|
46 |
+
self.histories = {} # 各 ID 分数历史队列
|
47 |
+
self.track_ids = set() # 所有被跟踪的ID
|
48 |
+
|
49 |
+
# 视频处理状态
|
50 |
+
self.cap = None
|
51 |
+
self.writer = None
|
52 |
+
self.frame_id = 0
|
53 |
+
self.results = [] # 存储每帧结果
|
54 |
+
self.start_frame = 0
|
55 |
+
self.end_frame = 0
|
56 |
+
|
57 |
+
def init_video(self, video_path, output_path=None, start_frame=0, end_frame=None):
|
58 |
+
"""初始化视频处理"""
|
59 |
+
# 打开视频并初始化写出器
|
60 |
+
self.cap = cv2.VideoCapture(video_path)
|
61 |
+
if not self.cap.isOpened():
|
62 |
+
raise IOError(f"无法打开视频 {video_path}")
|
63 |
+
|
64 |
+
# 获取视频属性
|
65 |
+
width = int(self.cap.get(cv2.CAP_PROP_FRAME_WIDTH))
|
66 |
+
height = int(self.cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
|
67 |
+
fps = self.cap.get(cv2.CAP_PROP_FPS) or 30
|
68 |
+
self.fps = max(fps, 1.0) # 保存帧率到实例变量,确保至少为1
|
69 |
+
total_frames = int(self.cap.get(cv2.CAP_PROP_FRAME_COUNT))
|
70 |
+
|
71 |
+
if self.verbose:
|
72 |
+
print(f"视频尺寸: {width}x{height}, 帧率: {fps}, 总帧数: {total_frames}")
|
73 |
+
|
74 |
+
# 设置帧范围
|
75 |
+
self.start_frame = start_frame
|
76 |
+
self.end_frame = end_frame if end_frame is not None else total_frames - 1
|
77 |
+
|
78 |
+
# 确保帧范围有效
|
79 |
+
if self.start_frame < 0:
|
80 |
+
self.start_frame = 0
|
81 |
+
if self.end_frame >= total_frames:
|
82 |
+
self.end_frame = total_frames - 1
|
83 |
+
if self.start_frame > self.end_frame:
|
84 |
+
self.start_frame, self.end_frame = self.end_frame, self.start_frame
|
85 |
+
|
86 |
+
# 将视频定位到起始帧
|
87 |
+
if self.start_frame > 0:
|
88 |
+
self.cap.set(cv2.CAP_PROP_POS_FRAMES, self.start_frame)
|
89 |
+
|
90 |
+
# 如果输出为视频则初始化 VideoWriter
|
91 |
+
if output_path and output_path.lower().endswith(('.mp4', '.avi')):
|
92 |
+
# 使用标准编码器
|
93 |
+
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
|
94 |
+
# 创建VideoWriter
|
95 |
+
self.writer = cv2.VideoWriter(output_path, fourcc, fps, (width, height))
|
96 |
+
if self.writer.isOpened():
|
97 |
+
print(f"成功创建输出视频: {output_path}, 尺寸: {width}x{height}")
|
98 |
+
else:
|
99 |
+
print(f"警告: 无法创建输出视频 {output_path}")
|
100 |
+
|
101 |
+
# 重置状态
|
102 |
+
self.frame_id = self.start_frame
|
103 |
+
self.results = []
|
104 |
+
self.prev_masks.clear()
|
105 |
+
self.histories.clear()
|
106 |
+
self.track_ids.clear()
|
107 |
+
|
108 |
+
if self.verbose:
|
109 |
+
print(f"视频初始化完成: 总帧数 {total_frames}, 分析范围 {self.start_frame}-{self.end_frame}")
|
110 |
+
|
111 |
+
return total_frames, self.start_frame, self.end_frame
|
112 |
+
|
113 |
+
def process_frame(self, frame, frame_id):
|
114 |
+
"""处理单帧,返回可视化帧和本帧结果列表"""
|
115 |
+
if self.verbose and frame_id % 10 == 0:
|
116 |
+
print(f"process_frame: 处理帧 {frame_id}")
|
117 |
+
|
118 |
+
try:
|
119 |
+
# 使用YOLO模型跟踪对象
|
120 |
+
results = self.model.track(
|
121 |
+
frame,
|
122 |
+
persist=True, # 保持跟踪ID的持久性
|
123 |
+
conf=self.conf,
|
124 |
+
iou=self.iou,
|
125 |
+
max_det=self.max_det,
|
126 |
+
verbose=False
|
127 |
+
)
|
128 |
+
|
129 |
+
# 检查是否有检测结果
|
130 |
+
frame_results = []
|
131 |
+
|
132 |
+
if results[0].boxes is None or len(results[0].boxes) == 0:
|
133 |
+
if self.verbose and frame_id % 50 == 0:
|
134 |
+
print("没有检测到任何对象")
|
135 |
+
return frame.copy(), []
|
136 |
+
|
137 |
+
# 处理检测结果
|
138 |
+
if hasattr(results[0], 'masks') and results[0].masks is not None:
|
139 |
+
# 获取掩码和跟踪ID
|
140 |
+
masks = results[0].masks.data.cpu().numpy()
|
141 |
+
track_ids = results[0].boxes.id
|
142 |
+
|
143 |
+
if track_ids is None:
|
144 |
+
if self.verbose and frame_id % 50 == 0:
|
145 |
+
print("没有获取到跟踪ID")
|
146 |
+
return frame.copy(), []
|
147 |
+
|
148 |
+
track_ids = track_ids.int().cpu().numpy()
|
149 |
+
|
150 |
+
if self.verbose and frame_id % 50 == 0:
|
151 |
+
print(f"检测到 {len(masks)} 个掩码,{len(track_ids)} 个跟踪ID")
|
152 |
+
|
153 |
+
# 更新跟踪ID集合
|
154 |
+
for track_id in track_ids:
|
155 |
+
self.track_ids.add(int(track_id))
|
156 |
+
|
157 |
+
# 处理每个跟踪对象
|
158 |
+
for i, (mask, track_id) in enumerate(zip(masks, track_ids)):
|
159 |
+
track_id = int(track_id)
|
160 |
+
|
161 |
+
# 二值化掩码
|
162 |
+
bin_mask = (mask > 0.2).astype(np.uint8)
|
163 |
+
|
164 |
+
# 应用形态学操作清理掩码
|
165 |
+
kernel = np.ones((5,5), np.uint8)
|
166 |
+
bin_mask = cv2.morphologyEx(bin_mask, cv2.MORPH_CLOSE, kernel)
|
167 |
+
|
168 |
+
# 调整掩码尺寸到与原始帧相同
|
169 |
+
if bin_mask.shape != (frame.shape[0], frame.shape[1]):
|
170 |
+
bin_mask = cv2.resize(bin_mask, (frame.shape[1], frame.shape[0]), interpolation=cv2.INTER_NEAREST)
|
171 |
+
|
172 |
+
# 计算挣扎度
|
173 |
+
if track_id in self.prev_masks:
|
174 |
+
prev_mask = self.prev_masks[track_id]
|
175 |
+
# 确保比较的掩码尺寸一致
|
176 |
+
if prev_mask.shape != bin_mask.shape:
|
177 |
+
prev_mask = cv2.resize(prev_mask, (bin_mask.shape[1], bin_mask.shape[0]), interpolation=cv2.INTER_NEAREST)
|
178 |
+
inter = np.logical_and(prev_mask > 0, bin_mask > 0).sum()
|
179 |
+
union = np.logical_or(prev_mask > 0, bin_mask > 0).sum()
|
180 |
+
iou = inter / union if union > 0 else 0
|
181 |
+
score = 1 - iou
|
182 |
+
if self.verbose and frame_id % 50 == 0:
|
183 |
+
print(f"跟踪ID {track_id} 挣扎分数: {score:.4f} (IoU: {iou:.4f})")
|
184 |
+
else:
|
185 |
+
score = 0.0
|
186 |
+
if self.verbose and frame_id % 50 == 0:
|
187 |
+
print(f"跟踪ID {track_id} 初始帧,分数为0")
|
188 |
+
|
189 |
+
# 保存当前掩码和历史
|
190 |
+
self.prev_masks[track_id] = bin_mask
|
191 |
+
|
192 |
+
if track_id not in self.histories:
|
193 |
+
self.histories[track_id] = collections.deque(maxlen=self.history_size)
|
194 |
+
self.histories[track_id].append(score)
|
195 |
+
|
196 |
+
# 计算挣扎状态
|
197 |
+
is_struggling = score >= self.struggle_threshold
|
198 |
+
|
199 |
+
# 计算质心
|
200 |
+
ys, xs = np.where(bin_mask > 0)
|
201 |
+
if len(xs) > 0:
|
202 |
+
centroid = (int(xs.mean()), int(ys.mean()))
|
203 |
+
else:
|
204 |
+
# 如果掩码为空,使用边界框中心点
|
205 |
+
box = results[0].boxes[i].xyxy.cpu().numpy()[0]
|
206 |
+
centroid = (int((box[0] + box[2]) / 2), int((box[1] + box[3]) / 2))
|
207 |
+
|
208 |
+
# 添加到帧结果
|
209 |
+
frame_results.append({
|
210 |
+
'id': track_id,
|
211 |
+
'score': float(score),
|
212 |
+
'centroid': centroid,
|
213 |
+
'is_struggling': is_struggling
|
214 |
+
})
|
215 |
+
else:
|
216 |
+
if self.verbose and frame_id % 50 == 0:
|
217 |
+
print("没有检测到任何掩码")
|
218 |
+
return frame.copy(), []
|
219 |
+
|
220 |
+
# 可视化 - 在这里创建最终的标注帧
|
221 |
+
annotated = frame.copy()
|
222 |
+
|
223 |
+
# 绘制掩码和ID
|
224 |
+
for result in frame_results:
|
225 |
+
track_id = result['id']
|
226 |
+
color = self.colors[track_id % len(self.colors)]
|
227 |
+
|
228 |
+
# 绘制掩码
|
229 |
+
if track_id in self.prev_masks:
|
230 |
+
mask = self.prev_masks[track_id]
|
231 |
+
# 确保掩码与帧大小一致
|
232 |
+
if mask.shape != (frame.shape[0], frame.shape[1]):
|
233 |
+
mask = cv2.resize(mask, (frame.shape[1], frame.shape[0]), interpolation=cv2.INTER_NEAREST)
|
234 |
+
mask_overlay = np.zeros_like(frame)
|
235 |
+
mask_overlay[mask > 0] = color
|
236 |
+
|
237 |
+
# 使用更精确的掩码边缘
|
238 |
+
contours, _ = cv2.findContours(mask.astype(np.uint8), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
|
239 |
+
cv2.drawContours(annotated, contours, -1, color, 2)
|
240 |
+
|
241 |
+
# 使用addWeighted进行混合
|
242 |
+
cv2.addWeighted(annotated, 1.0, mask_overlay, 0.4, 0, annotated)
|
243 |
+
|
244 |
+
# 在质心位置绘制ID和挣扎状态
|
245 |
+
centroid = result['centroid']
|
246 |
+
status_text = "Struggle" if result['is_struggling'] else "Static"
|
247 |
+
cv2.putText(annotated, f"ID:{track_id} {status_text}",
|
248 |
+
(centroid[0], centroid[1]),
|
249 |
+
cv2.FONT_HERSHEY_SIMPLEX, 0.7, color, 2)
|
250 |
+
|
251 |
+
# 在顶部创建黑色半透明条,显示总结信息
|
252 |
+
cv2.rectangle(annotated, (0, 0), (frame.shape[1], 40), (0, 0, 0), -1)
|
253 |
+
|
254 |
+
# 计算挣扎中的老鼠数量
|
255 |
+
struggling_count = sum(1 for r in frame_results if r['is_struggling'])
|
256 |
+
total_count = len(frame_results)
|
257 |
+
|
258 |
+
# 显示统计信息
|
259 |
+
cv2.putText(annotated, f"Total: {total_count} Struggling: {struggling_count}",
|
260 |
+
(10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2)
|
261 |
+
|
262 |
+
# 最后,由于OpenCV以BGR格式工作,但可能需要RGB格式,
|
263 |
+
# 确保返回的图像是BGR格式(视频写入用BGR,显示用RGB)
|
264 |
+
if annotated.dtype != np.uint8:
|
265 |
+
annotated = annotated.astype(np.uint8)
|
266 |
+
|
267 |
+
return annotated, frame_results
|
268 |
+
|
269 |
+
except Exception as e:
|
270 |
+
import traceback
|
271 |
+
if self.verbose:
|
272 |
+
print(f"处理帧时出错: {str(e)}")
|
273 |
+
traceback.print_exc()
|
274 |
+
# 返回原始帧和空结果
|
275 |
+
return frame.copy(), []
|
276 |
+
|
277 |
+
def process_video(self, video_path, output_path=None, start_frame=0, end_frame=None, callback=None):
|
278 |
+
"""处理整段视频,可选的回调函数用于更新进度"""
|
279 |
+
# 初始化视频
|
280 |
+
total_frames, start, end = self.init_video(video_path, output_path, start_frame, end_frame)
|
281 |
+
self.results = [] # 确保结果列表被清空
|
282 |
+
|
283 |
+
frame_id = start
|
284 |
+
processed_frames = 0
|
285 |
+
frames_to_process = end - start + 1
|
286 |
+
last_progress = -1
|
287 |
+
|
288 |
+
# 临时保存一帧,用于调试
|
289 |
+
debug_frame_saved = False
|
290 |
+
|
291 |
+
while frame_id <= end:
|
292 |
+
ret, frame = self.cap.read()
|
293 |
+
if not ret:
|
294 |
+
break
|
295 |
+
|
296 |
+
# 处理当前帧
|
297 |
+
annotated, frame_res = self.process_frame(frame, frame_id)
|
298 |
+
self.results.append(frame_res) # 将当前帧结果存入results列表
|
299 |
+
|
300 |
+
# 保存第一帧用于调试
|
301 |
+
if not debug_frame_saved and len(frame_res) > 0:
|
302 |
+
debug_frame_path = os.path.join(os.path.dirname(output_path), "debug_frame.jpg")
|
303 |
+
cv2.imwrite(debug_frame_path, annotated)
|
304 |
+
print(f"调试: 保存了标注帧到 {debug_frame_path}")
|
305 |
+
debug_frame_saved = True
|
306 |
+
|
307 |
+
# 写入输出视频
|
308 |
+
if self.writer:
|
309 |
+
# 确保帧是BGR格式
|
310 |
+
if len(annotated.shape) == 3 and annotated.shape[2] == 3:
|
311 |
+
# 如果需要,将RGB转换回BGR (OpenCV使用BGR)
|
312 |
+
# 默认应该已经是BGR,但为了确保
|
313 |
+
if frame_id == start:
|
314 |
+
print(f"调试: 写入标注帧到视频,形状: {annotated.shape}")
|
315 |
+
|
316 |
+
try:
|
317 |
+
self.writer.write(annotated)
|
318 |
+
except Exception as e:
|
319 |
+
print(f"调试: 写入帧到视频时出错: {str(e)}")
|
320 |
+
import traceback
|
321 |
+
traceback.print_exc()
|
322 |
+
|
323 |
+
# 更新进度和回调
|
324 |
+
processed_frames += 1
|
325 |
+
progress = int(100 * processed_frames / frames_to_process)
|
326 |
+
|
327 |
+
if progress != last_progress and callback:
|
328 |
+
callback(progress, annotated, frame_res)
|
329 |
+
last_progress = progress
|
330 |
+
|
331 |
+
frame_id += 1
|
332 |
+
|
333 |
+
# 释放资源
|
334 |
+
self.cap.release()
|
335 |
+
if self.writer:
|
336 |
+
self.writer.release()
|
337 |
+
print(f"调试: 视频写入完成,保存到: {output_path}")
|
338 |
+
|
339 |
+
return self.results
|
340 |
+
|
341 |
+
def save_results(self, csv_path):
|
342 |
+
"""导出分析结果到 CSV"""
|
343 |
+
import csv
|
344 |
+
with open(csv_path, 'w', newline='') as f:
|
345 |
+
writer = csv.writer(f)
|
346 |
+
writer.writerow(['frame_id', 'mouse_id', 'score', 'is_struggling'])
|
347 |
+
for fid, frs in enumerate(self.results):
|
348 |
+
for fr in frs:
|
349 |
+
writer.writerow([
|
350 |
+
fid + self.start_frame,
|
351 |
+
fr['id'],
|
352 |
+
f"{fr['score']:.4f}",
|
353 |
+
1 if fr.get('is_struggling', False) else 0
|
354 |
+
])
|
355 |
+
|
356 |
+
def generate_time_series_plot(self, threshold=None):
|
357 |
+
"""生成时序图分析"""
|
358 |
+
try:
|
359 |
+
print(f"Starting to generate time series plot with {len(self.results)} frames of data")
|
360 |
+
|
361 |
+
if not self.results or len(self.results) < 10:
|
362 |
+
print("Not enough data for time series plot (need at least 10 frames)")
|
363 |
+
return None
|
364 |
+
|
365 |
+
# 使用传入的阈值或默认阈值
|
366 |
+
if threshold is None:
|
367 |
+
threshold = self.struggle_threshold
|
368 |
+
|
369 |
+
# 使用保存的帧率,确保不会出现除以零的情况
|
370 |
+
fps = getattr(self, 'fps', None)
|
371 |
+
if fps is None or fps <= 0:
|
372 |
+
fps = 30 # 使用默认帧率
|
373 |
+
print(f"Warning: Invalid frame rate detected, using default: {fps} fps")
|
374 |
+
else:
|
375 |
+
print(f"Using frame rate: {fps} fps")
|
376 |
+
|
377 |
+
# 处理数据
|
378 |
+
frames = []
|
379 |
+
mouse_data = {}
|
380 |
+
mouse_positions = {} # 用于存储每只老鼠的平均X坐标
|
381 |
+
|
382 |
+
for frame_id, frame_results in enumerate(self.results):
|
383 |
+
frames.append(frame_id + self.start_frame) # 使用真实帧号
|
384 |
+
for result in frame_results:
|
385 |
+
mouse_id = result['id']
|
386 |
+
if mouse_id not in mouse_data:
|
387 |
+
mouse_data[mouse_id] = {'frames': [], 'seconds': [], 'scores': [], 'struggling': []}
|
388 |
+
mouse_positions[mouse_id] = [] # 初始化X坐标列表
|
389 |
+
|
390 |
+
frame_num = frame_id + self.start_frame
|
391 |
+
second = frame_num / fps # 转换为秒
|
392 |
+
|
393 |
+
mouse_data[mouse_id]['frames'].append(frame_num)
|
394 |
+
mouse_data[mouse_id]['seconds'].append(second)
|
395 |
+
mouse_data[mouse_id]['scores'].append(result['score'])
|
396 |
+
mouse_data[mouse_id]['struggling'].append(1 if result.get('is_struggling', False) else 0)
|
397 |
+
|
398 |
+
# 记录质心的X坐标
|
399 |
+
if 'centroid' in result:
|
400 |
+
mouse_positions[mouse_id].append(result['centroid'][0])
|
401 |
+
|
402 |
+
print(f"Processed data for {len(mouse_data)} mice")
|
403 |
+
if not mouse_data:
|
404 |
+
print("No valid mouse data to plot")
|
405 |
+
return None
|
406 |
+
|
407 |
+
# 计算每只老鼠的平均X坐标并按从左到右排序
|
408 |
+
avg_positions = {}
|
409 |
+
for mouse_id, positions in mouse_positions.items():
|
410 |
+
if positions:
|
411 |
+
avg_positions[mouse_id] = sum(positions) / len(positions)
|
412 |
+
else:
|
413 |
+
avg_positions[mouse_id] = float('inf') # 如果没有位置数据,放到最后
|
414 |
+
|
415 |
+
# 按从左到右排序老鼠ID
|
416 |
+
sorted_mice = sorted(mouse_data.keys(), key=lambda mid: avg_positions.get(mid, float('inf')))
|
417 |
+
print(f"Mice sorted from left to right: {sorted_mice}")
|
418 |
+
|
419 |
+
# 对数据进行平滑处理
|
420 |
+
def smooth_data(data, window_size=5):
|
421 |
+
"""使用移动平均平滑数据"""
|
422 |
+
if len(data) < window_size:
|
423 |
+
return data
|
424 |
+
smoothed = []
|
425 |
+
for i in range(len(data)):
|
426 |
+
start = max(0, i - window_size // 2)
|
427 |
+
end = min(len(data), i + window_size // 2 + 1)
|
428 |
+
window = data[start:end]
|
429 |
+
smoothed.append(sum(window) / len(window))
|
430 |
+
return smoothed
|
431 |
+
|
432 |
+
# 创建子图
|
433 |
+
num_mice = len(mouse_data)
|
434 |
+
fig, axes = plt.subplots(num_mice, 1, figsize=(12, 4*num_mice), sharex=True)
|
435 |
+
|
436 |
+
# 如果只有一只鼠,确保axes是列表
|
437 |
+
if num_mice == 1:
|
438 |
+
axes = [axes]
|
439 |
+
|
440 |
+
# 绘制每只老鼠的挣扎得分曲线,按从左到右的顺序
|
441 |
+
for idx, mouse_id in enumerate(sorted_mice):
|
442 |
+
data = mouse_data[mouse_id]
|
443 |
+
ax = axes[idx]
|
444 |
+
|
445 |
+
# 平滑数据
|
446 |
+
smoothed_scores = smooth_data(data['scores'], window_size=5)
|
447 |
+
|
448 |
+
# 绘制曲线
|
449 |
+
ax.plot(data['seconds'], smoothed_scores, label=f"Smoothed", color='blue', linewidth=2)
|
450 |
+
ax.plot(data['seconds'], data['scores'], label=f"Raw", color='lightblue', alpha=0.5, linewidth=1)
|
451 |
+
|
452 |
+
# 标记挣扎区域
|
453 |
+
for i, is_struggling in enumerate(data['struggling']):
|
454 |
+
if is_struggling:
|
455 |
+
ax.axvspan(data['seconds'][i]-0.5/fps, data['seconds'][i]+0.5/fps, alpha=0.1, color='red')
|
456 |
+
|
457 |
+
# 绘制阈值线
|
458 |
+
ax.axhline(y=threshold, color='r', linestyle='--', label=f"Threshold ({threshold:.2f})")
|
459 |
+
|
460 |
+
# 设置图表
|
461 |
+
ax.set_ylabel('Struggle Score')
|
462 |
+
position_text = f"(Position: Left #{sorted_mice.index(mouse_id)+1})" if mouse_id in avg_positions else ""
|
463 |
+
ax.set_title(f'Mouse {mouse_id} Struggle Score {position_text}')
|
464 |
+
ax.legend(loc='upper right')
|
465 |
+
ax.grid(True)
|
466 |
+
|
467 |
+
# 设置Y轴范围0-1
|
468 |
+
ax.set_ylim(-0.05, 1.05)
|
469 |
+
|
470 |
+
# 设置共享的X轴标签
|
471 |
+
axes[-1].set_xlabel('Time (seconds)')
|
472 |
+
|
473 |
+
# 动态调整x轴范围,精确到0.1秒
|
474 |
+
if frames:
|
475 |
+
start_time = self.start_frame / fps
|
476 |
+
end_time = max(frames) / fps
|
477 |
+
# 扩展一点范围以便更好地显示
|
478 |
+
axes[-1].set_xlim(start_time, end_time)
|
479 |
+
|
480 |
+
# 设置次要刻度(细网格线)
|
481 |
+
tick_interval = 0.1 # 保持0.1秒的细网格
|
482 |
+
minor_ticks = np.arange(start_time, end_time + tick_interval, tick_interval)
|
483 |
+
axes[-1].set_xticks(minor_ticks, minor=True)
|
484 |
+
|
485 |
+
# 设置主要刻度(标签和粗网格线)- 整秒
|
486 |
+
major_start = math.ceil(start_time)
|
487 |
+
major_end = math.floor(end_time)
|
488 |
+
major_ticks = np.arange(major_start, major_end + 1, 1.0) # 整秒刻度
|
489 |
+
axes[-1].set_xticks(major_ticks)
|
490 |
+
axes[-1].set_xticklabels([f"{int(t)}" for t in major_ticks]) # 整数秒标签
|
491 |
+
|
492 |
+
# 设置网格
|
493 |
+
axes[-1].grid(True, which='both')
|
494 |
+
axes[-1].grid(which='minor', alpha=0.2)
|
495 |
+
axes[-1].grid(which='major', alpha=0.5)
|
496 |
+
|
497 |
+
plt.tight_layout()
|
498 |
+
|
499 |
+
# 保存图表到临时文件并返回路径
|
500 |
+
temp_file = tempfile.NamedTemporaryFile(suffix='.png', delete=False)
|
501 |
+
plt.savefig(temp_file.name, dpi=150, bbox_inches='tight')
|
502 |
+
plt.close()
|
503 |
+
|
504 |
+
print(f"Time series plot saved to: {temp_file.name}")
|
505 |
+
return temp_file.name
|
506 |
+
|
507 |
+
except Exception as e:
|
508 |
+
import traceback
|
509 |
+
print(f"Error generating time series plot: {str(e)}")
|
510 |
+
traceback.print_exc()
|
511 |
+
return None
|
512 |
+
|
513 |
+
if __name__ == "__main__":
|
514 |
+
import argparse
|
515 |
+
|
516 |
+
parser = argparse.ArgumentParser(description="鼠强迫游泳实验挣扎度分析")
|
517 |
+
parser.add_argument('--video', type=str, required=True, help='输入视频路径')
|
518 |
+
parser.add_argument('--model', type=str, required=True, help='模型文件路径')
|
519 |
+
parser.add_argument('--output', type=str, help='输出视频路径')
|
520 |
+
parser.add_argument('--csv', type=str, help='输出CSV结果路径')
|
521 |
+
parser.add_argument('--conf', type=float, default=0.25, help='置信度阈值')
|
522 |
+
parser.add_argument('--iou', type=float, default=0.45, help='IOU阈值')
|
523 |
+
parser.add_argument('--max-det', type=int, default=20, help='最大检测数量')
|
524 |
+
parser.add_argument('--threshold', type=float, default=0.3, help='挣扎阈值')
|
525 |
+
parser.add_argument('--start', type=int, default=0, help='起始帧')
|
526 |
+
parser.add_argument('--end', type=int, default=None, help='结束帧')
|
527 |
+
parser.add_argument('--verbose', action='store_true', help='详细输出')
|
528 |
+
|
529 |
+
args = parser.parse_args()
|
530 |
+
|
531 |
+
# 设置输出路径
|
532 |
+
if not args.output:
|
533 |
+
video_name = os.path.splitext(os.path.basename(args.video))[0]
|
534 |
+
args.output = os.path.join(os.path.dirname(args.video), f"{video_name}_out.mp4")
|
535 |
+
|
536 |
+
if not args.csv:
|
537 |
+
video_name = os.path.splitext(os.path.basename(args.video))[0]
|
538 |
+
args.csv = os.path.join(os.path.dirname(args.video), f"{video_name}_results.csv")
|
539 |
+
|
540 |
+
# 创建分析器并处理
|
541 |
+
analyzer = MouseTrackerAnalyzer(
|
542 |
+
model_path=args.model,
|
543 |
+
conf=args.conf,
|
544 |
+
iou=args.iou,
|
545 |
+
max_det=args.max_det,
|
546 |
+
verbose=args.verbose
|
547 |
+
)
|
548 |
+
analyzer.struggle_threshold = args.threshold
|
549 |
+
|
550 |
+
# 进度回调函数
|
551 |
+
def progress_callback(progress, frame, results):
|
552 |
+
print(f"处理进度: {progress}%, 检测到 {len(results)} 个对象")
|
553 |
+
|
554 |
+
# 处理视频
|
555 |
+
analyzer.process_video(
|
556 |
+
video_path=args.video,
|
557 |
+
output_path=args.output,
|
558 |
+
start_frame=args.start,
|
559 |
+
end_frame=args.end,
|
560 |
+
callback=progress_callback
|
561 |
+
)
|
562 |
+
|
563 |
+
# 保存结果
|
564 |
+
analyzer.save_results(args.csv)
|
565 |
+
|
566 |
+
# 生成分析图表
|
567 |
+
plot_path = analyzer.generate_time_series_plot()
|
568 |
+
if plot_path:
|
569 |
+
print(f"挣扎度时序分析图已保存到: {plot_path}")
|
570 |
+
|
571 |
+
print(f"分析完成,视频已保存到: {args.output}")
|
572 |
+
print(f"结果数据已保存到: {args.csv}")
|