YOLO Vision Code
YOLO 实时人体检测：基础显示版
打开摄像头后直接运行 YOLOv8 人体检测，并在 OpenCV 窗口显示检测框与 FPS。
这些脚本用于 DRF450 或机载计算机上的 OpenCV + YOLOv8 视觉识别测试。建议先在桌面或树莓派本地验证摄像头索引、模型路径和性能，再把检测结果接入 MAVLink、自主飞行或任务触发逻辑。
脚本目标
读取摄像头画面，使用 yolov8n.pt 检测 COCO person 类别，实时绘制人体框、置信度和 FPS。
运行依赖： Python、OpenCV、Ultralytics YOLO。检测脚本默认模型路径为
/home/drobotics/yolov8n.pt，部署时请确认模型文件存在。学习重点
- 适合桌面环境调试模型和摄像头
- 保留完整画面尺寸进行推理
- 按 q 键退出，便于现场演示
关键参数
| 参数 | 当前设置 |
|---|---|
CAMERA_INDEX | 0 |
CONF_THRESHOLD | 0.5 |
PERSON_CLASS_ID | 0 |
运行前检查
- 安装依赖：
pip install opencv-python ultralytics - 确认摄像头编号，必要时修改
CAMERA_INDEX或CAMERA_INDEXES。 - 确认 YOLO 模型路径，例如
/home/drobotics/yolov8n.pt。 - 在无人机上运行前，先单独验证摄像头、推理速度、保存路径和散热。
完整代码：show_video.py
import cv2
from ultralytics import YOLO
import time
# =========================
# SETTINGS
# =========================
CAMERA_INDEX = 0
CONF_THRESHOLD = 0.5
# =========================
# LOAD MODEL
# =========================
print(" Loading YOLOv8 model...")
model = YOLO("/home/drobotics/yolov8n.pt") # small + fast for Raspberry Pi
# COCO class 0 = person
PERSON_CLASS_ID = 0
# =========================
# CAMERA INIT
# =========================
cap = cv2.VideoCapture(CAMERA_INDEX)
if not cap.isOpened():
print(" Cannot open camera")
exit()
print(" Camera started... Press 'q' to quit")
prev_time = 0
# =========================
# MAIN LOOP
# =========================
while True:
ret, frame = cap.read()
if not ret:
print(" Failed to read frame")
break
# YOLO inference
results = model(frame, verbose=False)[0]
# Draw detections
for box in results.boxes:
cls_id = int(box.cls[0])
conf = float(box.conf[0])
if cls_id == PERSON_CLASS_ID and conf > CONF_THRESHOLD:
x1, y1, x2, y2 = map(int, box.xyxy[0])
# draw box
cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 255, 0), 2)
label = f"Person {conf:.2f}"
cv2.putText(frame, label, (x1, y1 - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.5,
(0, 255, 0), 2)
# FPS calculation
curr_time = time.time()
fps = 1 / (curr_time - prev_time) if prev_time else 0
prev_time = curr_time
cv2.putText(frame, f"FPS: {fps:.2f}", (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 1,
(255, 0, 0), 2)
# show frame
cv2.imshow("Human Detection (YOLOv8)", frame)
# quit key
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
下一步
先完成摄像头采集，再运行实时检测或无界面检测。后续可以把人体检测结果接入 MAVLink 任务逻辑，实现识别触发、悬停、返航、录像或地面站告警。