Detecting an empty circular/rectangular on a surface using a pointcloud - c++

Question

I am using the Intel Realsense L515 lidar to obtain a pointcloud. I need to detect an empty area (no objects blocking it) on a surface (i.e. a table or a wall) so that a robotic arm has enough room to apply pressure to the surface at that point.
So far I have segmented the pointcloud in PCL in order to find a plane, to first detect the surface.
Afterwards I try segmenting for a 10-20cm circle anywhere in the area in order to attempt to find an empty space for the robot to operate.
https://imgur.com/a/fbY2yJ9
However as you can see, the circle (yellow points) is hollow instead of a full disc (so there could be objects lying inside it) and it has holes in it so it's not really an empty place on the table as the holes are created by objects eliminated in the planar segmentation stage (in this case that's my foot in the middle of the circle).
Is there any way to obtain the true non-object cluttered area on a planar surface?
Current code:
//Plane segmentation processing start
pcl::ModelCoefficients::Ptr coefficients (new pcl::ModelCoefficients);
pcl::PointIndices::Ptr inliers (new pcl::PointIndices);
// Create the plane segmentation object
pcl::SACSegmentation<pcl::PointXYZ> seg;
// Optional
seg.setOptimizeCoefficients (true);
// Mandatory
seg.setModelType (pcl::SACMODEL_PLANE);
seg.setMethodType (pcl::SAC_RANSAC);
seg.setDistanceThreshold (0.005);
// Create the filtering object
pcl::ExtractIndices<pcl::PointXYZ> extract;
// Segment the largest planar component from the remaining cloud
seg.setInputCloud(cloud);
pcl::ScopeTime scopeTime("Test loop plane");
{
seg.segment(*inliers, *coefficients);
}
if (inliers->indices.size() == 0)
{
std::cerr << "Could not estimate a planar model for the given dataset." << std::endl;
}
// Extract the inliers
extract.setInputCloud(cloud);
extract.setIndices(inliers);
extract.setNegative(false);
extract.filter(*cloud_p);
std::cerr << "PointCloud representing the planar component: " << cloud_p->width * cloud_p->height << " data points." << std::endl;
//Circle segmentation processing start
pcl::ModelCoefficients::Ptr coefficients_c (new pcl::ModelCoefficients);
pcl::PointIndices::Ptr inliers_c (new pcl::PointIndices);
// Create the circle segmentation object
pcl::SACSegmentation<pcl::PointXYZ> seg_c;
// Optional
seg_c.setOptimizeCoefficients (true);
// Mandatory
seg_c.setModelType (pcl::SACMODEL_CIRCLE2D);
seg_c.setMethodType (pcl::SAC_RANSAC);
seg_c.setDistanceThreshold (0.01);
seg_c.setRadiusLimits(0.1,0.2);
// Create the filtering object
pcl::ExtractIndices<pcl::PointXYZ> extract_c;
// Segment a circle component from the remaining cloud
seg_c.setInputCloud(cloud_p);
pcl::ScopeTime scopeTime2("Test loop circle");
{
seg_c.segment(*inliers_c, *coefficients_c);
}
if (inliers_c->indices.size() == 0)
{
std::cerr << "Could not estimate a circle model for the given dataset." << std::endl;
}
std::cerr << "Circle coefficients: \nCenter coordinates: " << coefficients_c->values[0] << " " << coefficients_c->values[1] << " " << coefficients_c->values[2] << " ";
// Extract the inliers
extract_c.setInputCloud(cloud_p);
extract_c.setIndices(inliers_c);
extract_c.setNegative(false);
extract_c.filter(*cloud_c);
std::cerr << "PointCloud representing the circle component: " << cloud_c->width * cloud_c->height << " data points." << std::endl;