Supported Data Formats

Detekt supports a variety of data types to facilitate comprehensive geospatial analysis. To ensure optimal performance and accurate results, users must adhere to the specified data requirements outlined below:

Images

  • Camera Models: Various supported (spherical, idealized cube, planar)
  • Additional Cameras: Recommended for enhanced results (e.g., pavement cameras for road damages)
  • Format: JPG
  • Orientation Information in CSV:some text
    • GPS Time: Timestamp associated with the image.
    • Positioning: Geographic coordinates (x, y, z).
    • Pitch, Yaw, and Roll: The three complex parameters crucial for establishing optimal orientation.

Image Metadata

  • Format: Ideally CSV for position and rotation
  • Intrinsic Camera Parameters: Can be provided in .txt or .ort (only required for planar cameras)
  • [Optional] Shapefiles: In case a CSV cannot be provided

CSV Structure

Filename | Timestamp | X | Y | Z | Pan | Tilt | Roll

Point Clouds

  • Format: LAS/LAZ
  • Mandatory: GPS time
  • Reference System: Same as image positions

Minimum Camera Setup

Basic Requirements (regardless of the use case)

  • Georeferenced position and orientation
  • Camera calibration: For planar cameras, the intrinsic parameters are needed (refer to Image Metadata
  • Lighting, environment: Ideal conditions include bright lighting and a high sun position, avoiding times close to sunrise or sunset. It is also important to ensure there is no rain or snow during data capture.
  • Car speed during capturing: less than 50 km/h (to ensure images are free from motion blur)

Image quality

  • JPEG compression: ~ 90%
  • No motion blur
  • Balanced exposure (ensuring that images are neither too bright nor too dark, with clear details in both shadows and bright areas)
  • No excessive noise

Use Case specific Sensor Requirements

Road Signs
Road Markings
Road Damages
Panoramic Camera*
~
x
Frame Camera
~
Pavement Camera
x
Lidar

Legend

Model is supported at its best capacity

~ Model is supported for only high severity detections (e.g., severe cracks)

x Model is not supported for the use case without the specific sensor (e.g., pavement camera for road damages)

Frame and Pavement Camera Requirements

Minimum
Typical
Optimal
Resolution
8 MP
24 MP
> 24 MP
Capturing Distance
< 5 meters
< 3 meters
1 meter
# Cameras
1
2
4
Field of View
-
72°
90°

Additional information regarding pavement cameras

  • Commonly used pavement camera systems include butterfly or rear cameras, for example, two planar cameras with a 60° field of view
  • Sensor size is not necessarily a major factor; for instance, Leica Pegasus butterfly cameras have only 14x10mm sensors
  • Wide-angle fisheye lenses, such as those used in GoPros, are not suitable as pavement cameras, as their spherical distortions impair pavement analysis