3D cameras are widely used nowadays, making possible 3D image reconstruction (for example, creating a 3D model of a room or a monument), and 3D estimation of a scene (of particular relevance for robotics and automotive applications). Several technologies build 3D models of a scene from images. Photogrammetry does 3D reconstruction by capturing many different images from different angles. NeRF (Neural Radiance Field) is based on the same principle, but uses AI to reduce the number of images required.

Stereovision estimates 3D from a calibrated set of two cameras. Some “active” systems include a projector to add texture to flat areas. Time-of-flight (TOF) sensors are also commonly used for 3D estimation. The smooth functioning of image-based 3D estimation depends greatly on precise calibration of intrinsic and extrinsic parameters, and on optical defects of the cameras. More generally, very good image quality is important for constructing an accurate 3D model. For example, a high signal-to-noise ratio (SNR) is required for a good image matching. 3D reconstruction devices often require working in various environments, from low light to sunlight, and in scenes with high dynamics.

Our testing solution includes a comprehensive hardware and software protocol for testing 3D estimation devices, based on ISO and IEEE standards. DXOMARK Analyzer makes it possible to measure all important image quality factors with great precision, including noise, dynamic range, optical parameters. The geometry bench allows for precise calibration and performance measurement of stereo cameras. DXOMARK also provides 3D targets as well as realistic mannequins for testing the quality of 3D models and 3D reconstruction. The multispectral lighting system (MLS) adjusts the impact of lighting on active and passive systems.

Use Cases:
  • Photogrammetry and NeRF
  • Active and passive stereo
  • Time-of-Flight (ToF)
  • Structured light
  • SLAM
Measurement and KPIs:
  • 3D geometry: calibration of intrinsic and extrinsicparameters for stereovision cameras
  • Precise distortion calibration (Brown-Conradymodels compatible with OpenCV and Matlab) witherror map for accuracy evaluation
  • Distortion and chromatic aberration
  • Noise and SNR
  • Texture and sharpness
Starter Kit: