first commit

This commit is contained in:
grtsinry43 2026-01-17 02:33:38 +08:00
commit e2d8be0356
Signed by: grtsinry43
GPG Key ID: F3305FB3A978C934
2 changed files with 159 additions and 0 deletions

77
main.py Normal file
View File

@ -0,0 +1,77 @@
import cv2
from ultralytics import YOLO
import time
def detect_bikes_modern(video_path, output_path=None):
# 1. 加载模型
# yolov8n.pt 是 "Nano" 版本,速度最快,适合 CPU 运行
# 第一次运行时会自动从互联网下载该文件 (约 6MB)
print("正在加载 YOLOv8 模型...")
model = YOLO('yolov8n.pt')
# 2. 打开视频
cap = cv2.VideoCapture(video_path)
if not cap.isOpened():
print(f"错误: 无法打开视频 {video_path}")
return
# 获取视频参数
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
fps = cap.get(cv2.CAP_PROP_FPS)
# 3. 设置保存 (可选)
out = None
if output_path:
# 现代环境通常使用 mp4v 或 avc1
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
out = cv2.VideoWriter(output_path, fourcc, fps, (width, height))
print(f"结果将保存到: {output_path}")
print("开始检测... 按 'q' 退出")
while True:
ret, frame = cap.read()
if not ret:
break
# 4. 核心检测逻辑
# stream=True 让处理更流畅
# classes=[1] 意思是我们只检测 COCO 数据集中的第 1 类 (0是人, 1是自行车, 2是汽车...)
# conf=0.3 意思是置信度大于 0.3 才算检测到
results = model.predict(frame, conf=0.2, classes=[1], verbose=False)
# 5. 在图上画框
# result.plot() 会自动把检测框画在 frame 上
annotated_frame = results[0].plot()
# 显示计数
bike_count = len(results[0].boxes)
cv2.putText(annotated_frame, f"Bikes: {bike_count}", (20, 40),
cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)
# 显示画面
cv2.imshow("Modern Bike Detection", annotated_frame)
# 保存画面
if out:
out.write(annotated_frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
if out:
out.release()
cv2.destroyAllWindows()
print("处理完成。")
if __name__ == "__main__":
# 替换成你的视频路径
# 注意:这里我们不再需要那个 lowerFAR.xml 文件了,直接扔掉它
video_source = "intersection.avi"
output_file = "result_modern.mp4"
detect_bikes_modern(video_source, output_file)

82
main2.py Normal file
View File

@ -0,0 +1,82 @@
import cv2
import numpy as np
def run_optical_flow(video_path):
# 1. 打开视频
cap = cv2.VideoCapture(video_path)
if not cap.isOpened():
print(f"错误: 无法打开视频 {video_path}")
return
# 2. 读取第一帧并转换为灰度
ret, first_frame = cap.read()
if not ret:
print("错误: 视频无法读取")
return
prev_gray = cv2.cvtColor(first_frame, cv2.COLOR_BGR2GRAY)
# 3. 设置可视化参数
step = 16 # 步长:不需要每个像素都画箭头,每隔 16 个像素画一个
scale = 1 # 缩放因子:控制箭头的长度,如果箭头太短看不清,把这个改大 (例如 2 或 3)
color = (0, 255, 0) # 绿色箭头
print("'q' 退出...")
while True:
ret, frame = cap.read()
if not ret:
break
# 转换为灰度图
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# 4. 计算稠密光流 (Farneback 算法)
# flow 是一个 (h, w, 2) 的数组flow[..., 0] 是水平移动(dx)flow[..., 1] 是垂直移动(dy)
flow = cv2.calcOpticalFlowFarneback(prev_gray, gray, None,
pyr_scale=0.5, levels=3, winsize=15,
iterations=3, poly_n=5, poly_sigma=1.2, flags=0)
# 5. 可视化:在原图上画箭头
# 创建网格点 (只在 step 的倍数位置画)
h, w = gray.shape
y, x = np.mgrid[step / 2:h:step, step / 2:w:step].reshape(2, -1).astype(int)
# 获取这些网格点对应的 dx, dy
fx, fy = flow[y, x].T
# 创建画线的图层
vis_frame = frame.copy()
# 绘制线条
lines = np.vstack([x, y, x + fx * scale, y + fy * scale]).T.reshape(-1, 2, 2)
lines = np.int32(lines + 0.5)
# 遍历绘制箭头 (cv2.arrowedLine)
for (x1, y1), (x2, y2) in lines:
# 只有当运动幅度超过一定阈值才画,减少噪点 (对应 MATLAB 的 NoiseThreshold)
if np.sqrt((x2 - x1) ** 2 + (y2 - y1) ** 2) > 1:
cv2.arrowedLine(vis_frame, (x1, y1), (x2, y2), color, 1, tipLength=0.3)
# 6. 显示结果
cv2.imshow('Optical Flow (Dense)', vis_frame)
# 更新上一帧
prev_gray = gray
# 按 'q' 退出
if cv2.waitKey(20) & 0xFF == ord('q'): # waitKey(20) 控制播放速度
break
cap.release()
cv2.destroyAllWindows()
if __name__ == "__main__":
# 请确保目录下有 thumb.avi或者改成 0 使用摄像头测试
video_file = "thumb.avi"
# 如果没有视频文件,想用摄像头测试,取消下面这行的注释:
# video_file = 0
run_optical_flow(video_file)