上传文件至 /

这个提交包含在:
2025-09-18 14:37:01 +00:00
父节点 998aab4546
当前提交 c9cfc3d3bc
共有 4 个文件被更改,包括 249 次插入 和 0 次删除

111
export_data.py 普通文件
查看文件

@ -0,0 +1,111 @@
import csv
import json
import os
from datetime import datetime
from config import *
from utils import create_directory, draw_detection
def export_detection_data(detection_data, filename):
"""导出检测数据到CSV文件"""
create_directory(EXPORT_DIR)
filepath = os.path.join(EXPORT_DIR, filename)
with open(filepath, 'w', newline='') as csvfile:
fieldnames = ['frame', 'timestamp', 'object_id', 'x', 'y', 'width', 'height']
writer = csv.DictWriter(csvfile, fieldnames=fieldnames)
writer.writeheader()
for data in detection_data:
writer.writerow(data)
print(f"检测数据已导出到: {filepath}")
def export_detection_summary(detection_data, filename):
"""导出检测摘要到JSON文件"""
create_directory(EXPORT_DIR)
# 计算摘要统计信息
total_frames = max([d['frame'] for d in detection_data]) if detection_data else 0
total_objects = len(detection_data)
unique_objects = len(set([d['object_id'] for d in detection_data]))
# 计算平均检测大小
avg_width = sum([d['width'] for d in detection_data]) / total_objects if total_objects > 0 else 0
avg_height = sum([d['height'] for d in detection_data]) / total_objects if total_objects > 0 else 0
summary = {
"export_date": datetime.now().isoformat(),
"total_frames": total_frames,
"total_detections": total_objects,
"unique_objects": unique_objects,
"average_width": avg_width,
"average_height": avg_height,
"detection_data": detection_data
}
filepath = os.path.join(EXPORT_DIR, filename)
with open(filepath, 'w') as jsonfile:
json.dump(summary, jsonfile, indent=4)
print(f"检测摘要已导出到: {filepath}")
def export_frames_with_detections(video_path, detection_data, output_dir):
"""导出包含检测结果的帧图像"""
create_directory(output_dir)
cap = cv2.VideoCapture(video_path)
if not cap.isOpened():
print(f"错误: 无法打开视频文件 {video_path}")
return
# 获取检测帧的集合
detection_frames = set([d['frame'] for d in detection_data])
frame_count = 0
saved_count = 0
while True:
ret, frame = cap.read()
if not ret:
break
frame_count += 1
if frame_count in detection_frames:
# 绘制检测框
frame_detections = [d for d in detection_data if d['frame'] == frame_count]
for detection in frame_detections:
frame = draw_detection(
frame,
detection['x'],
detection['y'],
detection['width'],
detection['height'],
f"Obj_{detection['object_id']}"
)
# 保存帧
filename = os.path.join(output_dir, f"frame_{frame_count}.jpg")
cv2.imwrite(filename, frame, [cv2.IMWRITE_JPEG_QUALITY, EXPORT_IMAGE_QUALITY])
saved_count += 1
cap.release()
print(f"已导出 {saved_count} 个包含检测结果的帧到 {output_dir}")
if __name__ == "__main__":
# 示例用法
sample_data = [
{"frame": 1, "timestamp": 0.1, "object_id": 0, "x": 100, "y": 100, "width": 50, "height": 50},
{"frame": 2, "timestamp": 0.2, "object_id": 0, "x": 105, "y": 105, "width": 55, "height": 55},
{"frame": 2, "timestamp": 0.2, "object_id": 1, "x": 200, "y": 200, "width": 60, "height": 60},
]
export_detection_data(sample_data, "sample_detection.csv")
export_detection_summary(sample_data, "sample_summary.json")

87
realtime_detection.py 普通文件
查看文件

@ -0,0 +1,87 @@
import cv2
import time
from config import *
from utils import draw_detection, preprocess_image
from export_data import export_detection_data
def realtime_detection():
"""实时摄像头检测"""
# 加载训练好的分类器
cascade = cv2.CascadeClassifier(MODEL_PATH)
if cascade.empty():
print("错误: 无法加载分类器文件")
return
# 打开摄像头
cap = cv2.VideoCapture(0)
if not cap.isOpened():
print("错误: 无法访问摄像头")
return
print("实时检测已启动,按 'q' 退出,按 's' 保存当前帧")
detection_data = []
frame_count = 0
while True:
ret, frame = cap.read()
if not ret:
break
frame_count += 1
timestamp = time.time()
# 预处理图像
gray = preprocess_image(frame)
# 检测物体
objects = cascade.detectMultiScale(
gray,
scaleFactor=SCALE_FACTOR,
minNeighbors=MIN_NEIGHBORS,
minSize=MIN_SIZE,
maxSize=MAX_SIZE
)
# 绘制检测结果
for i, (x, y, w, h) in enumerate(objects):
frame = draw_detection(frame, x, y, w, h, "Object")
# 记录检测数据
detection_data.append({
"frame": frame_count,
"timestamp": timestamp,
"object_id": i,
"x": x,
"y": y,
"width": w,
"height": h
})
# 显示结果
cv2.imshow('Real-time Object Detection', frame)
# 键盘操作
key = cv2.waitKey(1) & 0xFF
if key == ord('q'):
break
elif key == ord('s'):
# 保存当前帧
filename = f"exports/frame_{frame_count}.jpg"
cv2.imwrite(filename, frame)
print(f"已保存: {filename}")
# 释放资源
cap.release()
cv2.destroyAllWindows()
# 导出检测数据
export_detection_data(detection_data, "realtime_detection.csv")
print("检测数据已导出到 exports/realtime_detection.csv")
if __name__ == "__main__":
realtime_detection()

2
requirements.txt 普通文件
查看文件

@ -0,0 +1,2 @@
opencv-python>=4.5.0
numpy>=1.19.0

49
train_model.py 普通文件
查看文件

@ -0,0 +1,49 @@
import cv2
import os
import subprocess
from config import *
from utils import create_directory
def create_samples():
"""创建正样本向量文件"""
create_directory("models")
cmd = [
"opencv_createsamples",
"-info", "positive.txt",
"-num", str(NUM_POSITIVE_SAMPLES),
"-w", str(OBJECT_WIDTH),
"-h", str(OBJECT_HEIGHT),
"-vec", "positives.vec"
]
print("执行命令:", " ".join(cmd))
subprocess.run(cmd)
print("正样本向量文件已创建: positives.vec")
def train_cascade():
"""训练级联分类器"""
cmd = [
"opencv_traincascade",
"-data", "models",
"-vec", "positives.vec",
"-bg", "negatives.txt",
"-numPos", str(NUM_POSITIVE_SAMPLES // 2), # 使用一半的正样本
"-numNeg", str(NUM_NEGATIVE_SAMPLES // 2), # 使用一半的负样本
"-numStages", str(NUM_STAGES),
"-w", str(OBJECT_WIDTH),
"-h", str(OBJECT_HEIGHT),
"-featureType", "HAAR"
]
print("执行命令:", " ".join(cmd))
print("训练可能需要很长时间,请耐心等待...")
subprocess.run(cmd)
print("训练完成! 模型保存在 models/cascade.xml")
if __name__ == "__main__":
create_samples()
train_cascade()