Apply online to license this technology
Friend or foe (FOF) identification is an important aspect of combat operations for the military. An object, when first observed by an operator, may be very small in size, have poor contrast, and the brightness level of the object may be too bright or too dark to make an immediate and accurate identification. The operator can make one of two responses: (1) fire a weapon at the object feeling that it is hostile, or (2) passively wait with the presumption that the object may be friendly. Error, friendly fire (shooting the object when it is a friendly object) or not shooting the hostile object have significant consequences to the operator. Thus, both speed and accuracy are important performance attributes in this task.
To aid in the quick identification of objects in a digital image, Air Force scientists have further perfected computing algorithms based on voters (classifiers) common in commercial and research applications. In the software developed by the Air Force, the smallest number of voters to completely describe the information channel is selected such that computational advantages of simplicity and shorter calculation time may be obtained. The application demonstrates the smallest number of the independent set of voters that can render a decision using information-theoretic means.
In this schema, if the voters that identify the object as friend or foe each have a classification error less than 0.5 and all classifiers have the same error rate (an assumption), for an odd number of voters (classifiers), N, the correct decision rate increases with increasing N. Thus the error in a decision process can be made arbitrarily small if a large number of voters N can be used which operate on small error rates. The primary inputs are three variables:
- DR, is an information distance metric
- I(x; y) is a mutual information measure
- EF is an efficiency normalization measure
These three variables result in three votes (0 or 1) as to the friend or foe status of an object. A summation of votes of more than 1.5 results in a decision as to the identity of the object as a foe.
- A taxonomy for software voting algorithms used in safety-critical systems IEEE Transactions on Reliability ( Volume: 53, Issue: 3, Sept. 2004)
- Faster and simpler classification of objects
- US patent 8,027,537 available for license