We are seeking reliable and cost-effective methods for object detection and classification applicable to the field of autonomous driving.

Background:

Achieving reliable detection and classification with a minimum latency and computational resource need is a high challenge. A novel method of detection would provide multiple benefits.

Constraints:

  • Object type: car, truck, emergency vehicle and motor cycle
  • First detection range (first time vehicle appears in sensor range): day: <= 160m, night: <= 140m
  • range rate (max. approaching vehicle speed): day: 33m/s, night: 20m/s
  • True positive: 0-60m > 99%, true negative: 60-160m > 97%
  • False negative: 0-60m <1%, false positive: 0-160m < 1%
  • Range error: 0m – 45m < 6%, 45m – 90m < 8% , > 90m: < 9%
  • Low or no incremental cost when compared with existing solutions.

Desired outcome of the solution:

An innovative method of detecting and classifying/identifying objects with minimal latency and low computational resource.

Field of use and intended applications:

Automotive