Shooting tips

Shooting for orthomosaic generation

Choose walls with detectable features

• Plain white or featureless surfaces make SfM feature extraction difficult
• Textured walls — tile joints, windows, piping, etc. — work best

SfM runs on visible images

CRITIR uses only the visible images for SfM. Infrared images are used exclusively for temperature analysis, not for shape reconstruction. This means the quality of the visible images (focus, motion blur, exposure) directly determines orthomosaic quality.

Do not rename files

• CRITIR pairs infrared and visible images using capture timestamps and filename patterns
• Renaming or moving files between folders can break pairing
• Keep the original filenames from the time of capture

---

Drone photography

Overlap

• Aim for 70–80% or more both forward and side overlap
• Insufficient overlap directly reduces the SfM reconstruction rate

Flight pattern

• Fly a grid pattern — horizontal and vertical passes — while facing the wall head-on
• Choose a flight speed that keeps shutter speed conditions free of motion blur

Shooting angle

• The further the camera deviates from the wall normal, the more reflections and projection distortion increase
• Shoot as perpendicular to the wall as possible
• A large share of oblique images degrades orthomosaic quality

Never zoom

• Zooming changes the field of view, which can reduce SfM accuracy and break alignment with the infrared images
• Always shoot at 1x (no zoom)

---

Ground-based photography

Distance and angle

• Face the wall as squarely as possible (the more oblique the view, the more distortion and reflection)

Overlap

• Keep about 30–50% overlap with adjacent images
• Move a little at a time — left/right and up/down — between shots
• A tripod is recommended (prevents blur)

---

Combining drone and ground photography

Combining a drone with a ground camera makes it easier to cover the entire wall, but it also makes SfM harder.

Why SfM tends to fail

• Drone and ground cameras have different intrinsic parameters (focal length, distortion), so SfM cannot treat them as the same camera model
• Large differences in shooting distance create a large scale gap between images, making feature matching difficult
• Drones shoot from above while ground cameras shoot from below, so the viewpoint change is large and correspondences are hard to find even with overlap

Tips for combined capture

• Secure a generous transition zone: make the drone coverage and the ground coverage overlap vertically by 30% or more
• Shoot the transition zone at similar distances: matching fails when the scale gap is large, so keep the drone's wall distance and the ground camera's wall distance close within the transition zone
• Face the wall head-on: when combining, it is especially important that both sets of images are captured perpendicular to the wall

If it does not work

If the SfM reconstruction rate is low, try the following.

• Capture additional images in the transition zone (increase the image count)
• Generate the orthomosaic from only one source — drone or ground — and analyze the other as individual images
• Split the wall into separate sets (upper and lower sections) and generate a separate orthomosaic for each

---

Environmental conditions for infrared photography

Temperature difference matters most

The most important factor in infrared inspection is not the absolute air temperature but a sufficient temperature difference between the target and its surroundings.

• As a rule of thumb, a temperature difference of 10°C or more is desirable, though the required sensitivity depends on what you are detecting (delamination, water leakage, etc.)
• Shoot during periods when the temperature difference is stable (no rapid changes)

Time of day

• To avoid solar radiation effects, early morning (around sunrise) or after sunset is ideal
• Walls exposed to direct sunlight develop solar-induced unevenness, so wait long enough after the sun stops hitting the wall (several hours or more, depending on the material)
• Night photography also works if a sufficient temperature difference is present

Weather

• Rain: generally unusable — wetting, evaporative cooling, and reflections make measurements unstable
• Overcast: reduces solar unevenness, which is a plus, but unusable if the temperature difference is insufficient. Can work depending on conditions
• After rain: wait until the wall dries (surface moisture disrupts the temperature distribution)

Wind

• Strong wind reduces contrast through convective heat dissipation
• For flight safety as well, fly the drone only in calm wind conditions
• A common conservative guideline is 5 m/s or less (depends on equipment and site conditions)

---

Distance to the wall

Distance from the wall is the most important parameter determining the infrared GSD (Ground Sample Distance — the real-world size of one pixel).

• For both drone and ground capture, work backwards from the target GSD to the standoff distance
• Example: a 640×512 infrared sensor (FOV 40°) at GSD 2 cm/px → about 35 m from the wall
• The result varies greatly with the model and lens (FOV), so calculate it in advance

---

Image quality settings

Emissivity and reflected temperature

• Set an emissivity appropriate for the target on the camera (concrete/tile: about 0.95)
• Use consistent settings within a survey so comparisons in the report remain valid

Temperature range

• The camera's temperature range setting at capture time does not affect analysis (R-JPEG stores per-pixel temperature data)
• However, setting an appropriate range makes on-site visual checks more efficient

---

Common mistakes