1.共享内存（python共享内存实际案例，传输opencv frame）

主进程程序

send.py

import cv2
import numpy as np
from multiprocessing import shared_memory, resource_tracker

cap = cv2.VideoCapture(0)
if not cap.isOpened():
    print("无法打开 RTSP 流，请检查地址、网络连接或 GStreamer 配置。")
else:
    # 创建共享内存
    shm_name = 'shared_frame'
    frame_shape = (360, 640, 3)  # 假设视频帧的大小为 480x640
    shm = shared_memory.SharedMemory(create=True, size=np.prod(frame_shape) * np.dtype(np.uint8).itemsize, name=shm_name)
    frame_buffer = np.ndarray(frame_shape, dtype=np.uint8, buffer=shm.buf)
    try:
        while True:
            ret, frame = cap.read()
            if ret:
                # 将frame写入共享内存
                frame = cv2.resize(frame, (640, 360))
                np.copyto(frame_buffer, frame)

                cv2.namedWindow("RTSP Stream", cv2.WINDOW_NORMAL)
                cv2.imshow("RTSP Stream", frame)
                if cv2.waitKey(1) & 0xFF == ord('q'):
                    break
            else:
                print("无法读取视频帧")
                break
    finally:
        cap.release()
        cv2.destroyAllWindows()
        shm.close()
        shm.unlink()

receive.py

import cv2
import numpy as np
from multiprocessing import shared_memory, resource_tracker

# 共享内存的名称
shm_name = 'shared_frame'
frame_shape = (360, 640, 3)  # 假设视频帧的大小为 480x640

# 连接到共享内存
shm = shared_memory.SharedMemory(name=shm_name)
# 注销资源跟踪器,避免receive异常终止，关闭共享内存
resource_tracker.unregister(shm._name, 'shared_memory')
frame_buffer = np.ndarray(frame_shape, dtype=np.uint8, buffer=shm.buf)


try:
    while True:
        # 从共享内存中读取frame
        frame = np.copy(frame_buffer)

        cv2.namedWindow("Shared Memory Frame", cv2.WINDOW_NORMAL)
        cv2.imshow("Shared Memory Frame", frame)
        if cv2.waitKey(1) & 0xFF == ord('q'):
            break
finally:
    cv2.destroyAllWindows()
    shm.close()

终端1启动

send.py

终端2启动 receive.py

ps：使用共享内存的方式，可以实现容器内和容器外程序的通讯，且延迟极低。通过该方式可以实现容器内做推理，容器外做GUI显示