SBIR-STTR Award

Multi aperture vision system coupled to neural network processors
Award last edited on: 10/31/2017

Sponsored Program
SBIR
Awarding Agency
DOD : MDA
Total Award Amount
$551,095
Award Phase
2
Solicitation Topic Code
-----

Principal Investigator
Roy E Williams

Company Information

Federal Electro-Optics Inc

1850 Poplar Crescent
Memphis, TN 38111
   (901) 605-2722
   N/A
   N/A
Location: Single
Congr. District: 09
County: Shelby

Phase I

Contract Number: N/A
Start Date: 00/00/00    Completed: 00/00/00
Phase I year
1992
Phase I Amount
$54,998
Nearly every optical system built to date has been modeled after the human eye. Signal processing techniques are also designed to emulate human responses. These human characteristics are imposed upon non-human systems such as missile trackers and robotic vision. Such systems designed to perform simple tasks do not require complex human characteristics, and could be modeled after less sophisticated creatures such as insects. Position determination could be accomplished by coupling a neural network to a multi-aperture vision system. The neural network, acting as the insect visual processor, would use a training set of data with specified inputs and outputs. Once the training set has established the response of the system, the network is considered "trained" and can be programmed to respond in a similar manner to other input function. The insect ommatida, or eyelets, will be constructed using gradient index lenses, fiber optics, and detectors. The system will be "trained" and then stimulated by arbitrary object positions to obtain its response.

Phase II

Contract Number: N/A
Start Date: 00/00/00    Completed: 00/00/00
Phase II year
1993
Phase II Amount
$496,097
Federal Electro-Optics, Inc. (FEO) proposes to develop a novel, effective tracking system utilizing neural network processing of the signals from a multi-aperture vision system. This approach, inspired by insects and other arthropods, represents a departure from conventional techniques which are modeled after complex human visual system. Some of the proposed task are extensions of the successful efforts of Phase I, while others are innovative improvements conceived from the study conducted in Phase I. the proposed tasks enhance the performance of the tracker by improving the neural network processing, the optical system and by simulation with realistic scene information including multiple targets and noise. A hardware prototype of the tracker will be implemented.