OpenCV 跟踪器：模型未在函数 init 中初始化

在视频的第一帧，我运行一个对象检测器，它返回对象的边界框，如下所示：

<type 'tuple'>: ((786, 1225), (726, 1217), (721, 1278), (782, 1288))

我想将此边界框作为初始边界框传递给跟踪器。但是，我收到以下错误：

OpenCV Error: Backtrace (The model is not initialized) in init, file /Users/jenkins/miniconda/1/x64/conda-bld/conda_1486588158526/work/opencv-3.1.0/build/opencv_contrib/modules/tracking/src/tracker.cpp, line 81
Traceback (most recent call last):
  File "/Users/mw/Documents/Code/motion_tracking/motion_tracking.py", line 49, in <module>
    tracker.init(frame, bbox)
cv2.error: /Users/jenkins/miniconda/1/x64/conda-bld/conda_1486588158526/work/opencv-3.1.0/build/opencv_contrib/modules/tracking/src/tracker.cpp:81: error: (-1) The model is not initialized in function init

框架形状为1080 x 1920我传递给跟踪器的值如下所示：

我不确定我发送边界框的顺序是否错误，或者我是否做错了什么。

tracker = cv2.Tracker_create("MIL")
init_once = False

while True:

    (grabbed, frame) = camera.read()

    if not grabbed:
        break

    symbols = scan(frame)

    for symbol in symbols:
        if not init_once:
            bbox = (float(symbol.location[0][0]), float(symbol.location[0][1]), float(symbol.location[2][0]), float(symbol.location[2][1]))
            tracker.init(frame, bbox)
            init_once = True
            break
        # draw_symbol(symbol, frame)

    _, newbox = tracker.update(frame)

    if _:
        top_left = (int(newbox[0]), int(newbox[1]))
        bottom_right = (int(newbox[0] + newbox[2]), int(newbox[1] + newbox[3]))
        cv2.rectangle(frame, top_left, bottom_right, (200, 0, 0))
        cv2.imshow("asd", frame)

    out.write(frame)

out.release()