US 11,756,287 B2
Systems and methods for feature extraction and artificial decision explainability
Christopher M. Pilcher, Dallas, TX (US); Christopher Harris, Dallas, TX (US); John R. Goulding, McKinney, TX (US); and William D. Weaver, Dallas, TX (US)
Assigned to Raytheon Company, Waltham, MA (US)
Filed by Raytheon Company, Waltham, MA (US)
Filed on Jan. 10, 2022, as Appl. No. 17/572,485.
Application 17/572,485 is a continuation of application No. 16/887,508, filed on May 29, 2020, granted, now 11,222,245.
Prior Publication US 2023/0244756 A1, Aug. 3, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06V 10/40 (2022.01); G06F 18/241 (2023.01)
CPC G06F 18/241 (2023.01) [G06V 10/40 (2022.01); G06V 2201/07 (2022.01)] 20 Claims
OG exemplary drawing
 
1. An automatic target recognizer system comprising:
a memory arranged to store target recognition data including a plurality of reference features associated with each of a plurality of reference targets;
a processor arranged to:
i) select a portion of the target recognition data based on a reference first gating feature of the plurality of reference features;
ii) process a first image received from an image acquisition system, the first image being associated with an acquired first target, wherein the preprocessor determines an acquired first gating feature of the acquired first target;
iii) discriminate the acquired first gating feature with the reference first gating feature and, if there is a match, extract a plurality of segments of the first image;
iv) detect one of the presence, absence, probability or likelihood of one of the plurality of features associated with each of the plurality of reference targets;
v) create a plurality of classification problems associated with the portion of the target recognition data based on the reference first gating feature;
vi) create a plurality of decision trees that solve the plurality of classification problems;
vii) select the decision tree having the minimal misclassification cost, iv) determine a classification of the acquired first target based on the decision tree having the minimal misclassification cost: and
viii) generate a classification decision report based on the determined classification of the acquired first target; and
a display arranged to display the classification decision report.