CityIntegrator

Video analytics

Applying the EN 62676-4 DORI Standard to Video Analytics and AI Surveillance

Detect, Observe, Recognize, and Identify: How to calculate exact PPM targets for AI models and prevent expensive camera re-rigging.

1 min read

Executive Briefing: Precision Optical Planning via European Standard EN 62676-4

System integrators frequently encounter project disputes when installed cameras fail to produce legally admissible facial identification or accurate neural analytics. Grounding camera specifications in the international EN 62676-4 DORI standard ensures that optical density (Pixels Per Meter / PPM) precisely matches operational requirements.

The Four Levels of DORI

  • Detection (25 PPM): Determines whether a human or vehicle is present within the frame. Insufficient for identification or classification.
  • Observation (62 PPM): Allows operators to view characteristic details such as clothing color or general vehicle type.
  • Recognition (125 PPM): Enables operators to determine with high confidence whether an individual has been seen before or is a known person of interest.
  • Identification (250 PPM): Guarantees sufficient facial feature detail to establish identity beyond reasonable doubt in legal court proceedings.

Adapting DORI for AI & Neural Analytics

Modern convolutional networks require specific minimum pixel densities to operate reliably. While human operators can recognize a familiar person at 125 PPM, deep learning facial biometrics engines require a minimum of 200 PPM across the inter-pupillary distance. Similarly, robust ANPR engines require at least 150 to 250 PPM across the license plate width to maintain 99%+ OCR accuracy.

Related articles

View all