Fieldwork: Planning a Scan

Good fieldwork starts with a recce - a reconnaissance of the site, in person where possible and otherwise from plans, photos and a conversation with whoever knows the place. You are gathering the facts that shape the plan: the size and layout, the access and any locked or restricted areas, the power and lighting, the likely occluders (furniture, plant, vegetation, parked vehicles), the tricky surfaces (glass, mirrors, polished stone, water), the safety hazards, and the practicalities of getting yourself and the kit in and around. A recce need not be elaborate - even a careful look at a floor plan and a few site photos transforms a blind arrival into a planned one - but the difference it makes is enormous, because it lets you plan coverage instead of improvising it.

Coverage planning is the core of the craft: deciding where the instrument will sit or fly so that every surface the brief needs is recorded, from enough angles, with the overlap that registration requires. For terrestrial scanning this means laying out stations - scan positions - across the space so their views overlap generously and between them they see into every nook. The enemy is occlusion: a scanner or camera only records what is in line of sight, so anything behind a column, inside a reveal, above a beam or around a corner is a shadow unless another position sees it. The figure shows the principle - one station leaves a shadow behind a column, and a second, deliberately placed station fills it. Plan the extra positions that kill the shadows; they are cheap on site and impossible to add later.

The same logic governs the other methods. Photogrammetry needs many overlapping photos - a generous, systematic set with high overlap between adjacent frames and views from enough angles that every surface appears in several images; thin or patchy coverage simply fails to reconstruct. Drone capture needs flight lines planned for overlap (forward and side) and the right height and angle for the subject, whether that is a roof, a terrain or a facade. Handheld and SLAM capture needs a planned, continuous path that keeps good overlap and avoids the featureless stretches where self-location drifts. Across all of them the principle is identical: plan positions and paths for complete, overlapping coverage with the occlusions deliberately filled, because coverage you did not plan is coverage you will discover missing when it is too late.

Coverage gets the surfaces recorded; control and registration planning makes the separate scans or images into one coherent, correctly-scaled, correctly-placed dataset - and deciding how that will happen is part of planning, not an afterthought on site. The question to answer in advance is: how will these scans tie together, and to what? Two layers matter. The first is registration - aligning the separate scans or image sets to each other into one consistent coordinate frame. The second, when the job needs it, is tying that frame to control: known reference points, and for survey-grade or georeferenced work, points of known real-world coordinates.

Different methods lean on different aids. Many scanning workflows register by overlap alone (cloud-to-cloud), which is why generous overlap in the coverage plan matters so much; others use targets - spheres or flat checkerboard markers placed in the scene and seen from multiple positions - to make registration robust and to check its quality, as the blue markers in the station figure suggest. Photogrammetry is scaled and oriented using known distances (scale bars) or control points, because images alone do not know real size. When a job must sit accurately in real-world coordinates, control points are surveyed in - and this is exactly where the work often crosses into the domain of a licensed surveyor. Plan, in advance, which aids you will use, how many, and where they go, so registration is designed rather than hoped for.

Two honest cautions belong here. First, registration error accumulates: chaining many scans or images can drift, so a plan that relies on a long daisy-chain of marginal overlaps will produce a cloud that looks crisp locally but is subtly distorted overall. Good planning builds in redundancy - generous overlap, well-distributed targets, and where possible closing loops - so the registration is well-constrained, and it plans to *check* the registration error, not assume it. Second, survey-grade georeferencing and control are not a casual add-on: tying a capture accurately to real-world coordinates needs proper control and qualified practice, and anything binding belongs to a licensed surveyor working to recognised standards (in India, within the Survey of India context). Plan the tie-together you are competent to do, design in the redundancy that keeps registration honest, and recognise the point at which control becomes a surveyor's responsibility rather than yours.

A plan that ignores the real conditions of the site meets reality badly. Lighting can make or break a capture: photogrammetry depends on good, even light and struggles in gloom, harsh shadow or changing sun, while laser scanning, being active, is far less fussy - so plan the timing and any lighting you need around the method. Weather rules outdoor and aerial work: wind grounds or destabilises drones, rain and fog degrade or stop most capture, and the low sun of early morning or late afternoon casts long shadows that hurt photogrammetry - so plan for the forecast and a fallback. Access is a logistics problem to solve before the day, not during it: keys and permissions, the path for heavy kit, power, the hours you are allowed on site, someone to let you in, and the sequence that lets you cover the whole place without backtracking or being locked out of a wing.

Safety is non-negotiable and comes first, always. Reality capture regularly puts people in demanding environments - live construction sites, heights, roofs, confined spaces, roads and rail corridors, operational buildings, heritage structures in poor repair. Plan the hazards out: the right protective equipment, permits to work, exclusion from traffic and machinery, safe access at height, awareness of live services, and never working alone where that is unsafe. For drones, safety merges with regulation - the national Drone Rules and DGCA framework govern where and how you may fly, and planning a flight means clearing the airspace and, often, engaging a licensed operator. No deliverable is worth an injury; if a position or a flight cannot be done safely and lawfully, it does not happen, and you plan an alternative.

Finally, plan the unglamorous but project-saving matter of data management on site. Captures generate large volumes of data fast, and the field is where it is most vulnerable. Plan enough storage and power - spare cards, charged and spare batteries - so you never have to cut coverage short. Establish a clear naming and organisation scheme as you go, so scans, photo sets and targets are identifiable later rather than an anonymous heap. And treat backup as a field task, not an office one: offload and back up data before you leave, ideally to more than one place, because a lost or corrupted card after you have left site means the whole field day is gone. The mantra is simple and hard-won: data that exists in only one place, on site, does not really exist yet. Plan the conditions, plan the safety, plan the data - and the field day survives contact with the real world.

Everything in this lesson converges on one habit, and if you take away nothing else, take this: before you leave the site, check that you have what you came for. This is the cheapest, most powerful quality control in the entire pipeline, because it exploits the one thing you will never have again - being on site with the equipment still set up. A gap found now costs minutes to fill; the identical gap found back at the office costs a return trip, a delayed project and your credibility. The field figure ends on exactly this gate: a deliberate 'can I leave?' moment, not a hopeful packing-up.

Run a concrete checklist, adapted to the method. Coverage: has every surface the brief requires been captured, from enough angles, with the occlusions filled? Walk the space against your plan and look specifically for the shadows - behind furniture, in reveals, above head height, around corners. Registration and overlap: will the scans actually tie together? Where the workflow allows a live or preview registration on site, check it; otherwise confirm that overlap and targets are sufficient and well-distributed, because discovering a registration that will not close is vastly cheaper now. Control: are the targets or control points recorded and identifiable? Data integrity: are all the files actually saved, offloaded and backed up, with nothing stranded on a single card, and is the kit (batteries, cards) accounted for? Only when those pass do you pack up.

This check closes the feedback loop from Lesson 7.1 at the cheapest possible point. QA is not a single event at the end of the project; it is threaded throughout, and the on-site check is its most valuable instance because it acts before the irreversible stage ends. It pairs with the on-site note that becomes part of the deliverable: what you captured, by what method, with what targets and control, and any known gaps or limits you could not resolve - honest documentation that lets the next stage trust the data and that flags anything needing a surveyor. Build the before-you-leave check into every field day as an unbreakable ritual, and you convert fieldwork from the scary irreversible stage into a controlled one. Plan the coverage, plan the tie-together, plan the conditions, safety and data, and verify before you go - that is the whole discipline of planning a scan, and it is what makes a capture you can actually build on.