OpenCV-Python重叠boundingRect()

Question

我正在尝试修复一个错误，即当两个矩形在视频中发生碰撞时，它们会合并为一个大矩形以用于碰撞的其余帧，而不是像以前一样保留矩形：

碰撞前：

碰撞后：

即使物体发生碰撞，我也想保留两个矩形。

主循环片段：

while True:
    ret, frame = vid.read()
    if frame is None:
        break

    # TRESH_OTSU has problems with objects that are very dark
    bright_frame = cv2.addWeighted(
        frame, 4.3, np.zeros(frame.shape, frame.dtype), 0, 20
    )

    # create a mask that will remove the black background from tracking
    gray = cv2.cvtColor(bright_frame, cv2.COLOR_BGR2GRAY)
    _, mask = cv2.threshold(gray, 1, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)

    # get the moving objects
    contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    detections = []
    for obj in contours:
        area = cv2.contourArea(obj)
        if area > 50:  # filter noise
            perimeter = cv2.arcLength(obj, True)
            approx = cv2.approxPolyDP(obj, 0.02 * perimeter, True)
            object_type = "rectangle" if len(approx) == 4 else "circle"
            x, y, w, h = cv2.boundingRect(obj)
            detections.append(
                DetectedObject(
                    x, y, w, h, get_obj_color(frame, x, y, w, h), object_type
                )
            )

    rects = tracker.update(detections)
    for rect in rects:
        x, y, w, h, id, _, _, time = rect
        cv2.rectangle(frame, (x, y), (x + w, y + h), (255, 255, 255), 5)
        # cv2.putText(
        #     frame, str(id), (x, y - 15), cv2.FONT_HERSHEY_PLAIN, 5, (255, 255, 255)
        # )
        print(rect)
    cv2.imshow("frame", frame)

    key = cv2.waitKey(10)

    # break the loop upon the press of esc
    if key == 27:
        break

Answer 1

这是我的意思的一个例子。我做了一个小例子，因为你没有提供未处理的图像：

分离：

碰撞：

方法非常简单：

1. 定义边界
2. 申请

   cv2.inRange

3. 获取正像素并获取边界矩形

对于失散的人：

对于碰撞：

代码片段：

# define boundary for first object
lower1 = (150,90,30)
upper1 = (170,110,60)
# define boundary for second object
lower2 = (150,150,150)
upper2 = (255,255,255)
# get bounding boxes
boundingRect1 = cv2.boundingRect(cv2.inRange(im,lower1,upper1))
boundingRect2 = cv2.boundingRect(cv2.inRange(im,lower2,upper2))
# draw rectangle on im
cv2.rectangle(im, boundingRect1, (255,0,0), 5)
cv2.rectangle(im, boundingRect2, (0,0,255), 5)

我不确定你最终会拥有多少个物体，我也不确定这两个物体是否可以按颜色分开。如果你总是有像照片这样干净的对象，最终看看连接的组件