I am running openvslam slam on my robot. I am intending to publish the pointcloud2 topic from this. But since visual slam only extracts the sparse piontcloud. I have used the pointcloudtolaser package to convert the dense pointcloud into 2d laserscan and than create a 2d occupency grid map using slam modules like gmapping and hector. How can i use the sparse point cloud generated from the visual slam like openvslam to do the same i.e create 2d occupency grid map?