Michael Karlesky

A cabinet of wonders. Minus the cabinet. And possibly the wonders.

High Precision Targeting Research

Can a single pixel captured in-flight yield precise ground coordinates in real time?


A view of the research system hardware.

Despite perception, advanced, long-range military targeting systems are not, in fact, particularly accurate. A Michigan company won a Naval Research Lab contract to develop an experimental system to translate — in real-time — pixels from live aircraft reconnaissance photos into accurate targeting coordinates. The system combined camera angle and focal length with high precision GPS and inertial guidance data to correlate pixels to specific location coordinates. I was contracted to create a real-time, in-flight system from the original post-processing system. It would take more than a year of work to determine success.


A labeled view of the research system hardware.

The concurrency, timing, data production, and data consumption of the system’s individual sub-systems were extraordinarily demanding. From the outset, it was entirely possible that a realtime system could not be done.

I assembled a high-speed PC/104 stack running QNX. I solved the real-time logging, filtering, data acquisition, and testing challenges by creating individual executables for each major software component of the system. Each executable possessed standardized mechanisms for publishing to and consuming data from the others. A single executor coordinated startup and the inter-process subscription process of the executables.


The system allowed extensive testing with simulators for the camera, GPS, and inertial guidance data on the ground before flight testing. Ultimately, it proved successful enough to secure further research funding and spawn derivative versions.

When the system was ported to another platform, military engineers informed my contract client that I had (unknowingly) implemented a textbook version of a particular military specification for real-time software.