Horizontal regression relies on clean, well-structured geometric data. Poorly prepared strings (polylines) or points can lead to incorrect alignment detection, unreliable curvature plots, and time-consuming rework.
Use the guidelines below to ensure your data produces accurate, repeatable results.
Strings (2D/3D Polylines and FeatureLines)
Goal: Provide a single, continuous baseline that represents the intended centerline (or rail, pipe, road edge, etc.) with uniform vertex spacing.
- Keep vertex spacing consistent. A 2 m - 10 m separation yields the most stable results.
- Maintain a single, unbroken sequence of vertices; ensure start-to-end direction matches design intent.
- Simplify geometry by converting arcs into short chords only if absolutely required by the workflow (convert 2D Polyline to 3D Polyline removes arcs).
Avoid:
- Duplicated vertices (two points with exactly the same coordinates).
- Dangling segments or off-shoots that do not belong to the main line.
- Overlaps or self-intersections.
- Vertex "clusters" (several points within a few millimetres of each other).
- Vertices closer than the minimum model tolerance: merge or delete them.
- Embedded true arcs in a polyline; convert to chorded segments first.
COGO Points
Goal: Provide an ordered set of discrete points that follows the alignment in sequence.
- Name or number points so alphabetical or numerical sorting reproduces the physical order along the alignment (e.g., 0001, 0002, 0003 …).
- Verify spacing: large gaps reduce accuracy, and extremely dense clusters increase noise.
- Filter obvious outliers before export.
Avoid:
- Overlapping points (identical coordinates).
- Points located within the survey tolerance of each other: merge or delete duplicates.
- Relying on software “order by creation time”; always enforce your own naming convention.
Summary
High-quality input data minimizes manual corrections and maximizes the success rate of automated horizontal regression. Follow the cleaning steps above, and validate with a quick visual review.