SBIR-STTR Award

Automatic Intelligent Tracking of Passenger Vehicles in Aerial Video Feeds from Medium Altitude UAV (e.g., MQ-9), using Cognitive Explainable-AI
Award last edited on: 10/10/2023

Sponsored Program
SBIR
Awarding Agency
DOD : AF
Total Award Amount
$1,999,400
Award Phase
2
Solicitation Topic Code
AF221-DCSO1
Principal Investigator
Saied Tadayon

Company Information

Z Advanced Computing Inc

11204 Albermyrtle Road
Potomac, MD 20854
   (301) 294-0434
   info@zadvancedcomputing.com
   www.zadvancedcomputing.com
Location: Single
Congr. District: 06
County: Montgomery

Phase I

Contract Number: 2022
Start Date: ----    Completed: 5/5/2022
Phase I year
2022
Phase I Amount
$1
Direct to Phase II

Phase II

Contract Number: FA8649-22-P-0870
Start Date: 2/9/2024    Completed: 5/5/2022
Phase II year
2022
(last award dollars: 1696969621)
Phase II Amount
$1,999,399

ZAC’s disruptive Cognitive Explainable-AI (artificial intelligence) detailed 3D image recognition technology will be adapted to intelligently track target objects of interest (e.g., passenger vehicles) automatically in real-time within aerial video feeds from UAV (e.g., MQ-9), based on the target’s detailed description (which is critical for Intelligent Tracking). This proposed solution will significantly enhance USAF ISR capability to automatically, accurately, and robustly track the target across viewing interruptions and varying orientations and scenery, within the UAV video feed, to hand off tracking to other UAV feeds, or to automatically search through vast amounts of archived feeds captured by persistent ISR, to provide decision quality results. ZAC’s breakthrough detailed image recognition, based on our Explainable-AI, is a demonstrated technology, capable of detecting fine details in images taken at various camera angles (3D), overcoming the major limitations of neural networks which fail to recognize details beyond very generic categories or classifications, and thus, fail to distinguish between similar vehicles. ZAC’s approach beat ResNet (a state-of-the-art Deep Convolutional Neural Network (CNN)) in recognizing fine details of complex 3D objects in images, where ResNet completely failed to learn such fine features during its training. In addition, ZAC breakthrough algorithms, as demonstrated in projects with Bosch and USAF, require only a few training samples (much smaller than other algorithms) and much lower computational power, making it practical for edge computing, e.g., on-board UAS, which is one of the mission needs of AFSOC/A5K (Unmanned ISR Requirements). ZAC has successfully completed a project for ZAC Locator with MQ-9 Program Office (AFLCMC/WIIZ, Medium Altitude UAS Special Projects Branch) to automatically locate objects (passenger vehicles) from aerial images. ZAC also has an ongoing project with the same program office for ZAC Describer for recognizing detailed description of passenger vehicles automatically, which is critical to distinguish between similar looking vehicles. The proposed solution, as one of its objectives, integrates both ZAC Locator and ZAC Describer with Intelligent Tracking to find, fix, and track a target vehicle within the same feed or other video feeds. The attached customer memorandum (MOU) from the same Program Office supports the current proposal. ZAC won “Engage-Space Challenge” 2020 competition by the US Air Force (AFWERX) for the space industry, as one of the 26 finalists, out of the original 800 teams (i.e., among about the top 3%). Letter of Support from AFRL Space Vehicles Directorate is included. ZAC was recognized as the Top 5 Leading Global Companies in the 4th Industrial Revolution, by Oxford Academic in 2021. ZAC team includes world-renowned researchers/advisors, including Prof. David Lee (Nobel Prize in Physics) and General Michael Moseley, USAF (ret), the USAF 18th Chief of Staff.