You may consider removing the upper and lower camera and do your own "Manual blending" with only the middle camera. Recently I've been using Cinemachine and I do something similar of what your desired result is.
Since I don't exactly know how do you want your camera to behave, I show you some of my manual blending I've done, explained:
//Camera Direction
//If I´m not in ground, and I've been on the air for a specific time
if (!onGround && timeSinceLastJump > cameraDelay)
{
//Moves the offset of the camera down to the maximum allowed y offset (In your case it would be where the lower camera is)
if (vcam.GetCinemachineComponent<CinemachineTransposer>().m_FollowOffset.y >= maxYOffset)
vcam.GetCinemachineComponent<CinemachineTransposer>().m_FollowOffset.y -= offsetYSensivity;
//It also zooms out up to a specified level
if (vcam.m_Lens.OrthographicSize < maxFOV)
vcam.m_Lens.OrthographicSize += camSensivity;
}
else
{
//Same but upwards
if (vcam.GetCinemachineComponent<CinemachineTransposer>().m_FollowOffset.y <= minYOffset)
vcam.GetCinemachineComponent<CinemachineTransposer>().m_FollowOffset.y += offsetYSensivity;
//Same but zooming in
if (vcam.m_Lens.OrthographicSize > minFOV)
vcam.m_Lens.OrthographicSize -= camSensivity;
}
This way you can use your player height in the conditions, at the cost of having to design well the camera logic.
Maybe this helps you in some way to do what you want.