US 12,444,166 B2
Object classification based on spatially discriminated parts
Suhail Shabbir Saquib, Shrewsbury, MA (US); Christopher M. Pilcher, Richardson, TX (US); and John R. Goulding, Farmersville, TX (US)
Assigned to Raytheon Company, Arlington, VA (US)
Filed by Raytheon Company, Arlington, VA (US)
Filed on May 27, 2022, as Appl. No. 17/827,096.
Prior Publication US 2023/0419640 A1, Dec. 28, 2023
Int. Cl. G06V 10/764 (2022.01); G06V 10/22 (2022.01); G06V 20/17 (2022.01)
CPC G06V 10/764 (2022.01) [G06V 10/235 (2022.01); G06V 20/17 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method for machine learning (ML) automatic target recognition (ATR) decision explanation, the method comprising: receiving an object specification matrix from an object model database that indicates, for each of a plurality of physical portions of an object, whether each of a plurality of features are present or absent in a physical portion of the physical portions of the object and a proportional physical displacement between the features in the object; receiving, from one or more deep learning (DL) models, feature data indicating for an image of a portion of the object, a likelihood whether each of the features are present in the image; determining, by a classifier and based on the object specification matrix and the feature data, classification results that includes respective probabilities per object that the image corresponds to the object, respective probabilities per class that the object corresponds to the class and whether features of the plurality of features are present or absent in the image; and providing, by a user interface, the classification results with the probabilities sorted in descending order and including (i) a user-specified number of the features or (ii) the features associated with probabilities that exceed a user-specified threshold.