Skip to content
Browse docs / PARADA

Vision & OCR

The plate pipeline and its published Phase 14 benchmark.

Pipeline

A classical OpenCV plate-region detector (blackhat → Sobel-x → Otsu → close → aspect-ratio contour filter) finds candidate plates; EasyOCR 1.7.2 reads them on the CPU. Camera sources can be USB, RTSP or a video file. A read counts as trusted at confidence ≥ 0.5.

Phase 14 evaluation

Measured on a seeded, synthetic dataset — 1,250 images: 50 plate texts under 23 conditions, plus 100 scenes with no plate. It characterises the pipeline; it is not a production accuracy claim.

ResultValue
Exact plate reads (misses count as wrong)77.4% (890 / 1,150)
Character accuracy on detected plates93.76%
False plates on empty scenes0 / 100
Precision of trusted reads85.0%
ConditionExact
------
Clean, lighting, sensor noise100%
Rotation and perspective98.5%
Gaussian and motion blur75.0%
Distance50.5%
Partial occlusion33.3%
Most misreads are truncations — a correct part of the plate read with high confidence — which is why the backend never trusts confidence alone. The full method and failure analysis live in the repository's docs/vision/phase14-evaluation.md.