How to detect holes by using depth camera?

edit

Hi All,

Currently, I am trying to use a depth camera to detect holes on the ground and avoiding them. When using it, I found that we can only set the min_obstacle_height and max_obstacle_height in costmap parameters. And in source code of navigation costmap, both ros2 and ros1 are using same algorithm to check whether a point cloud is an obstacle, which is comparing z value by <=max_obstacle_height and >=min_obstacle_height.
But one problem of this algorithm is if the ramp’s slope is smooth, it may return false positive obstacles.

My question is: is there any algorithm which is based on the slope of the points net? or an algorithm which is not truncating data based on a single parameter, but more dynamic.

Thanks.
Best,
Marshal