Map Products Are Decision Artifacts

A mapping product is useful because somebody intends to make a decision from it.

That means the product should be judged by fitness for that decision, not by how impressive it looks on a screen.

Different outputs answer different questions

Photogrammetry workflows can produce several kinds of outputs.

Orthophoto / orthomosaic

A top-down image product geometrically transformed so locations can be represented in a map-like view.

Useful for:

• visual inspection;
• plan-view comparison;
• annotation;
• some measurements when scale/accuracy are validated.

Point cloud

A collection of reconstructed 3D points.

Useful for:

• shape analysis;
• elevation/structure inspection;
• creating later surface products;
• identifying areas where reconstruction is sparse or noisy.

Digital surface/elevation products

Gridded height representations derived from the reconstruction.

Useful for:

• terrain/surface comparison;
• slope/elevation analysis;
• volume-related workflows when accuracy is established.

3D mesh / textured model

A connected surface representation, often with image texture.

Useful for:

• visualization;
• spatial context;
• some geometry inspection;
• communicating complex structures.

No output is automatically “best.” The stakeholder question decides which artifact is useful.

Orthomosaic does not mean raw photograph

A stitched orthomosaic is a processed product.

The workflow may involve:

• camera-pose estimation;
• geometric correction;
• projection;
• surface/elevation assumptions;
• seam selection;
• blending;
• resampling.

That means the final image is not simply one untouched photograph of the entire site.

Processing decisions can affect what appears where.

For important claims, preserve access to source images and processing metadata.

Point density is not the same as accuracy

A dense point cloud can contain millions of points.

That does not prove those points are correctly positioned.

Density tells you how much reconstructed sampling exists.

Accuracy depends on geometry, camera model, reference control, validation, and other factors.

A billion wrong points are still wrong.

Outputs inherit weak areas from the input network

Week 14 identified a low-overlap corner.

If the software fills that region poorly, several products may inherit the problem:

• orthomosaic may stretch or blur;
• point cloud may be sparse;
• elevation surface may contain noise or holes;
• mesh may bridge across missing geometry.

A stakeholder looking only at the final screenshot may never know.

That is why quality flags should travel with the product.

Worked stakeholder case: facilities manager

A facilities manager asks:

Which roof zones show visible storm damage, and where should a maintenance crew inspect first?

A useful product package might include:

• orthomosaic or plan-view image for location context;
• clearly labeled roof zones;
• source-image links for close visual evidence;
• annotations marking suspected damage;
• note of areas with poor reconstruction or obstruction;
• measurement values only where scale/accuracy are strong enough;
• a limitations section.

A 3D model may help communicate roof geometry but may not be the fastest artifact for the manager’s decision.

Product choice should reduce decision friction.

Annotation creates a new evidence layer

Suppose a student draws a red polygon around “storm damage.”

The red polygon is not sensor data.

It is an interpretation added by the analyst.

A good annotation record identifies:

• who created it;
• which source evidence supports it;
• whether the feature is confirmed or suspected;
• when it was added;
• what uncertainty remains.

Otherwise analyst interpretation can become visually indistinguishable from measured data.

Metadata is part of the product

A map without context can be dangerously easy to misuse.

Useful metadata can include:

• project/date;
• coordinate reference system where relevant;
• capture method;
• processing software/version;
• scale/control method;
• checkpoint or validation summary;
• known coverage defects;
• intended use;
• limitation statement.

The exact fields depend on the project.

The goal is to stop the artifact from being separated from the evidence needed to interpret it.

Worked misuse: the construction measurement

A fictional orthomosaic was created for visual progress documentation.

It has:

• strong visual coverage;
• consumer image geolocation;
• no independent ground control;
• no checkpoint accuracy test.

Someone later asks:

Can we use this to order steel based on a 4.237 m measured roof feature?

No—not from the current evidence.

The product may be excellent for visual documentation and still be unvalidated for fabrication-grade dimensional decisions.

That is not a failed map.

It is a misused map.

Build a product-to-decision matrix

Use a supplied project and compare:

Some stakeholder questions may be outside the validated use of the dataset. Mark them that way.

Misconception: one map should answer every question

Different questions require different evidence.

A visual orthomosaic, a validated survey product, a dense point cloud, and a textured 3D model are not interchangeable simply because they came from the same images.

The stakeholder needs the right artifact and the right confidence level.

The Week 15 starting point

Treat every mapping output as:

product + metadata + validation evidence + intended use + limitations.

Without those extra pieces, a polished image can travel farther than the evidence that made it trustworthy.