AI generated/Shutterstock image

Optical range measurements can be accomplished by active or passive imaging. Active approaches involve the emission of laser radiation by the observer, which might cause hazard to the surrounding vicinity and make the observer visible to others, compromising effective intelligence, surveillance, and reconnaissance (ISR) and sense and avoid (SAA) capabilities. Passive approaches are important because they do not emit radiation and cannot be detected or jammed.

DARPA seeks to explore the potential of novel computational imaging, machine learning (ML) algorithms, and artificial intelligence (AI) tools and techniques to enhance the accuracy of passive ranging for tactical and civil applications, such as augmented reality and autonomous vehicles.

The CIDAR challenge aims to discover passive imaging algorithms for high-accuracy, low-latency distance measurements that equal or exceed the performance of today’s active range measurement systems. Specifically, the challenge will extend passive range measurements to 10 km or more with high accuracy while minimizing floating point operations to achieve low latency. The Cramer-Rao bound defines the fundamental limit of distance information in images, but passive imaging approaches to range measurements today can only capture ~1% of this information. The accuracy of range measurement algorithms improves 10–100x when information from a single spatial, spectral, or temporal optical filter is added to the information in unfiltered images. If we create new algorithms that integrate information from all optical filters, then we may be able to increase the accuracy of passive range measurements 10x–100x further to approach the fundamental limit of distance information in images. If the challenge succeeds, active range measurements like laser detection and ranging (LADAR) and laser range finding (LRF) can be performed passively without sacrificing speed or accuracy.