结合face_recognition使用cv2.VideoCapture时视频窗口崩溃无响应问题

你遇到的崩溃是典型的OpenCV GUI事件循环阻塞问题：你当前使用cnn模型做人脸检测，CPU运行时单帧处理耗时极高，导致OpenCV无法通过waitKey及时处理窗口的刷新、交互事件，系统判定窗口无响应后强制结束，对应错误码就是0xCFFFFFFF。单独运行OpenCV摄像头捕获逻辑时没有高耗时计算，所以功能正常。

按优先级从高到低尝试：

• 优先更换人脸检测模型：将代码中的MODEL = "cnn"改为MODEL = "hog"。HOG模型是CPU优化的轻量检测模型，普通PC上能跑到30帧以上，完全满足日常人脸匹配需求，精度损失极小。
• 补充读帧合法性判断：原代码未判断video.read()的返回值，摄像头读取失败时后续处理空帧会触发未知异常，需要在拿到帧后先判断ret是否为True，失败就退出循环。
• 调整waitKey等待时长：将cv2.waitKey(1)改为cv2.waitKey(10)，给GUI事件循环留足够的处理时间，10ms的延迟人眼感知不到，不会影响视频流畅度。
• 更换稳定依赖版本：如果修改模型后还是崩溃，可尝试将依赖换成稳定组合：dlib==19.24、opencv-python==4.5.5.62、face-recognition==1.3.0，避免不同版本的兼容问题。
• （可选）启用GPU加速：如果你有NVIDIA显卡，编译安装支持CUDA的dlib版本，cnn模型就能通过GPU加速跑满帧率，不会再出现阻塞问题。

import face_recognition
import os
import cv2

video = cv2.VideoCapture(0)
# 补充摄像头打开校验
if not video.isOpened():
    raise IOError("无法打开摄像头")

KNOWN_FACES_DIR = "known_faces"
TOLERANCE = 0.6
FRAME_THICKNESS = 3
FONT_THICKNESS = 2
MODEL = "hog"  # 替换为hog模型
FONT = cv2.FONT_HERSHEY_SIMPLEX

print("loading known faces...")

known_faces = []
known_names = []

for name in os.listdir(KNOWN_FACES_DIR):
    for filename in os.listdir(f"{KNOWN_FACES_DIR}/{name}"):
        image = face_recognition.load_image_file(f"{KNOWN_FACES_DIR}/{name}/{filename}")
        encoding = face_recognition.face_encodings(image)[0]
        known_faces.append(encoding)
        known_names.append(name)

print("processing unknown faces...")

while True:
    ret, frame = video.read()
    # 补充读帧失败判断
    if not ret:
        print("读取摄像头帧失败，退出程序")
        break

    locations = face_recognition.face_locations(frame, model=MODEL)
    encodings = face_recognition.face_encodings(frame, locations)

    for face_encoding, face_location in zip(encodings, locations):
        results = face_recognition.compare_faces(known_faces, face_encoding, TOLERANCE)
        match = None
        if True in results:
            match = known_names[results.index(True)]
            print(f"match found! = {match}")

            top_left = (face_location[3], face_location[0])
            bottom_right = (face_location[1], face_location[2])

            color = [0, 255, 0]
            cv2.rectangle(frame, top_left, bottom_right, color, FRAME_THICKNESS)

            top_left = (face_location[3], face_location[2])
            bottom_right = (face_location[1], face_location[2]+22)
            cv2.rectangle(frame, top_left, bottom_right, color, cv2.FILLED)
            cv2.putText(frame, match, (face_location[3]+10, face_location[2]+15), FONT, 0.5, (0, 0, 0), FONT_THICKNESS)

    cv2.imshow('frame', frame)
    # 调整waitKey时长
    if cv2.waitKey(10) & 0xFF == ord('q'):
        break

video.release()
cv2.destroyAllWindows()
# 部分Windows环境下需要额外加几帧waitKey保证窗口正常销毁
for _ in range(5):
    cv2.waitKey(1)

内容的提问来源于stack exchange，提问作者Yusuf