今日输出: 使用opencv进行背景提取,并进行帧差分析

使用opencv-python进行背景提取

1. 读取视频文件

video_path = '自己的视频路径,avi格式'
capture = cv2.VideoCapture(video_path)

2.   构造形态学使用的 kernel ,构造高斯混合模型

se = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
model = cv2.createBackgroundSubtractorMOG2()  # history=500, varThreshold=16, detectShadows=True

3. 逐帧读取视频, 同时进行视频帧的对比,使用到 opencv 内置的函数

frameNum = 0
while True:
    ret, im0 = capture.read()
  
    # 运用高斯模型进行拟合，在两个标准差内设置为0，在两个标准差外设置为255
    fg_mk = model.apply(im0)
    ret1, binary_img = cv2.threshold(fg_mk.copy(), 220, 255, cv2.THRESH_BINARY)
    
    # 获取背景
    binary_img = cv2.morphologyEx(binary_img, cv2.MORPH_OPEN, se)
    bg_image = model.getBackgroundImage()

    frameNum += 1
    temp_frame = bg_image

    # 作帧差 每次与序号整除50的帧作对比
    if (frameNum % 50 == 0):
        previous_frame = cv2.cvtColor(temp_frame, cv2.COLOR_BGR2GRAY)
        # print(previous_frame.shape)
        # 500帧之后才进行对比
        # 下面这个if语句为测试代码 实际使用的时候需要删掉
        if frameNum == 500:
            previous_frame = np.random.randint(2555, size=(480, 720))
            print(previous_frame.shape)
    if (frameNum >= 1000):
        currentframe = cv2.cvtColor(temp_frame, cv2.COLOR_BGR2GRAY)
        diff_frame = cv2.absdiff(previous_frame, currentframe)
        # diff_frame[diff_frame > 0] = 255
        # 打印差值
        print(cv2.sumElems(currentframe)[0] - cv2.sumElems(previous_frame)[0])
        diff = cv2.sumElems(currentframe)[0] - cv2.sumElems(previous_frame)[0]
        # 若差值大于自己规定的阈值 那么就会打印信息
        if diff > 5000 or diff < -5000:
            print("warning !!! The camera is not in the original place !!!")
        # 若差值很小 则看起来 diff_frame 就是黑色的 因为大部分差值为0
        cv2.imshow('1', diff_frame)

    # cv2.imshow('1', bg_image)
    cv2.waitKey(1)
capture.release()