As a first answer, I would say no. The number of splays is not necessarily a good indicator, as there are at least two other important parameters, in my opinion: the 3D distribution of the splays and the degree of splay overlap.
I suggested using the to and from shots as splays only when the concave envelopes do not overlap and do not completely contain the corresponding to and/or from shot(s). Otherwise, including them by default would unnecessarily affect the concave hull, even though the existing splay overlap is already sufficient to achieve a good envelope.
Plan: Every station is classified from its own splays and from how far they reach toward each neighbouring station:
LRUD — fewer than 3 splays. The station contributes only its LRUD cross-section. CONCAVE — its splays meet those of every neighbour: for each shot, the station's own reach plus the neighbour's reach covers at least 150 % of the shot length. The station gets its own concave hull. CONVEX — enough splays, but on at least one shot the two fans do not meet, so unmeasured space remains along that shot.
Reach is the largest projection of a splay onto the shot axis, divided by the shot length, counting only splays within 45° of the shot direction — a splay fired sideways does not cover the way to the neighbour, however long it is.
The geometry of a shot follows from the types of its two ends:
LRUD – LRUD classic LRUD tube CONCAVE – CONCAVE nothing: the two point hulls already overlap anything else one convex hull spanning the shot
Each CONCAVE station additionally emits its own concave hull — exactly once, no matter how many shots meet there.
For the convex shot hull, each end contributes differently: a CONVEX end gives all its splays, an LRUD end gives its LRUD cross-section as a rounded ring, and a CONCAVE end gives only the ring of splays roughly perpendicular to the shot (60°–120°).
It is definitely difficult to find a good balance between the different applicable models and merge them into a single algorithm.
The latest beta provides a reasonable solution for areas with a low splay density, but it significantly degrades the quality in areas with a high splay density.
beta:
vs latest version:
The current Scan model (not the one from the latest beta) performs quite well in areas with a high density of splays as well as with stations with LRUDs only or without LRUDs/splays (=tube). Where the Scan model struggles is in areas with insufficient splay coverage. Unfortunately, this lack of coverage is not homogeneous: sometimes there are simply too few splays, sometimes their spatial distribution is poor, sometimes the overlap between neighbouring stations is insufficient or completely missing and sometimes just a combination of all...
On the other hand, variations in splay density and spatial distribution are generally quite consistent within a given survey session, since they mainly depend on the surveyor's acquisition habits.
Wouldn't it therefore be more useful, within the Scan model, to keep the general model that way but to allow individual stations to use a different reconstruction model when needed? This would follow the same principle already used for LRUDs, where the behaviour can be overridden on a station-by-station basis through the LRUD column.
It's good that you mention the option to configure the hull model per station. I had already thought about implementing it, but put it aside because I wasn't sure whether it would be used in practice. My estimate for the implementation with Claude: 30 minutes.
CaveRenderPro
Thema drucken
30.07.2026 07:59
Antworten
