Photogrammetry

Photogrammetry reconstructs 3D shape from overlapping photographs. Cameras capture the hall, the yard, or the machine from many viewpoints. Software finds matching features across images, solves camera poses, and estimates depth. The dense result is a point cloud. Most plant projects then turn that cloud into a textured mesh people can walk in a viewer.

Facilities teams use photogrammetry when the CAD aisle no longer exists, when a contractor needs an as-built shell fast, or when a drone or phone pass is cheaper than a full survey campaign. The method answers the same plant question as LiDAR scanning: what does this hall actually look like now. The sensor path is different. Photogrammetry measures shape from image parallax. LiDAR measures range with a laser.

The output feeds a static digital twin and can become the spatial container for an operational digital twin later. It does not refresh when a forklift moves. A new photo set is required when the physical hall changes enough to mislead a clearance check or a training walk. A longer walkthrough of the pipeline lives in how photogrammetry builds plant 3D models from photos.

Key Components

Overlapping image set: Photos of the same surfaces from different angles, with enough overlap that features can be matched.

Camera poses: Where each photo was taken and which way it faced, solved so all views share one 3D frame.

Dense reconstruction: The point cloud built from stereo depth across the matched views.

Mesh and texture: Surfaces fitted to the cloud, with photo color applied so the hall looks familiar to people who walk it.

Scale and registration: Control points, measured distances, or a georeference so the model matches plant meters (and feet) and sits on the same origin as CAD or BIM.

Applications in Manufacturing and Logistics

Inside the plant, phone, tablet, or dedicated camera walks capture cells, mezzanines, and rack runs after a layout change. Drone passes cover large roofs, yards, and outdoor storage when a top-down model is needed. The mesh supports design reviews, contractor handoff, and training in a hall that matches reality better than an old drawing.

Logistics sites use the same method for dock faces, trailer lanes, and warehouse shells when the building file is thin. Once the mesh is trusted, live positions from RTLS or warehouse coordinates can sit on it. The photos built the map. They did not track the shift.

Benefits and Challenges

The benefit is accessible capture. Many plants already have cameras and drones. Textured meshes help supervisors recognize columns and doors without reading a coordinate. Cost per square meter (and per square foot) can be lower than a dense LiDAR campaign when lighting and access are good.

The challenge is lighting, texture, and scale. Dark, shiny, or featureless surfaces fail to match. Poor overlap leaves holes. Without measured control, the model can look right and still be the wrong size. Processing time and file size grow with hall area. Like any scan, the result freezes the day of capture.

Related Terms

Photogrammetry sits beside LiDAR scanning as a peer capture path. Both produce a point cloud and usually mesh geometry for a digital twin or static twin. 3D visualization is how people view the result. SLAM may also use cameras, but for live vehicle pose rather than a planned hall survey.

Frequently Asked Questions

Can phone photos build a usable plant model? Yes for many cells and smaller halls, if overlap, lighting, and scale control are handled. Large sites still often need planned camera or drone paths and measured control points.

Do we still need LiDAR if we have photogrammetry? Not always. Use photogrammetry when photo coverage is strong and texture matters. Use LiDAR when range must be measured directly or lighting is poor. Many twin projects combine both.

Is the mesh a live twin? No. The mesh is static geometry. Live forklifts and material states need an operational layer on top of that hall.

How do we keep the model current? Reshoot after layout changes that would mislead clearance or training. Daily material movement is not a reason to rebuild the whole mesh.