How does inside option of feature to point work?

[ChatterArcGISH]

Hi everyone!

I am interested in determining centroids for polygons. Some shapes result in the centroids to be outside the polygon if the "inside" option is not chosen. This makes sense as the geometric center might lay outside.

Now I need the created points to be within the shapes. Choosing the "inside" option however produces different points, even for those polygons where the centroid is within the polygon.

Can someone explain how the inside option determines the points that it gives as an output? I would've expect to receive centroids and only for those where centroids are outside the polygon different points, e.g. the closest point within the polygon to the centroid outside of it. But all of this is not the case and research on this has not been succesful.

[DanPatterson]

the "label point" algorithm revealed by esri (proprietary)

How To: Calculate Feature Centroids

The algorithms for the specific calculation are proprietary, but there are two basic concepts used.

the centroid of convex shapes uses well-known approaches, just not concave shapes

[ChatterArcGISH]

"Enabling this parameter forces the tool to calculate a centroid that is inside the feature boundaries, if it originally falls outside the boundaries the point will be adjusted to what is considered the center of gravity within the boundaries."

This sounds to me like it should give the same results as without the outside option as long as the real geometric centroid is within the polygon. As described in the original post, this is not the case. There are instances where the centroid function without the inside option yields a centroid that is within the polygon but the centroid function with the inside option yields a different centroid. And what is the center of gravity that is within the polygon? I would expect this to be the nearest point on the edge of the polygon to the outside geometric centroid. But this is not the case.

[DanPatterson]

Contact Tech Support,  perhaps someone from the appropriate development team can provide an explanation.

It would be interesting to see a detailed explanation of how the label point is derived