US 12,085,640 B2
Fuzzy labeling of low-level electromagnetic sensor data
Yihang Zhang, Calabasas, CA (US); Kanishka Tyagi, Agoura Hills, CA (US); and Narbik Manukian, Los Angeles, CA (US)
Assigned to Aptiv Technologies AG, Schaffhausen (CH)
Filed by Aptiv Technologies AG, Schaffhausen (CH)
Filed on Apr. 5, 2022, as Appl. No. 17/658,089.
Claims priority of provisional application 63/265,756, filed on Dec. 20, 2021.
Prior Publication US 2023/0194700 A1, Jun. 22, 2023
Int. Cl. G01S 13/90 (2006.01); G01S 7/41 (2006.01); G01S 13/86 (2006.01); G06N 3/043 (2023.01); G06N 3/08 (2023.01)
CPC G01S 13/9005 (2013.01) [G01S 7/417 (2013.01); G01S 13/865 (2013.01); G01S 13/867 (2013.01); G06N 3/043 (2023.01); G06N 3/08 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
identifying, based on first sensor data obtained from a first sensor, a geometric location of at least one object;
identifying, on a spectrum map derived from second sensor data obtained from a second sensor, an energy spectrum smear that corresponds to the object, the second sensor being an electromagnetic sensor;
identifying a first portion of the energy spectrum smear that corresponds to the geometric location of the object;
labeling, in the first portion of the energy spectrum smear, each pixel with a value of one;
labeling, in a second portion of the energy spectrum smear that includes all pixels of the energy spectrum smear not included in the first portion, each pixel with a value between zero and one, the value decreasing the further each respective pixel is from the first portion of the energy spectrum smear; and
training, by machine learning and based on the labeling of each pixel in the first portion and each pixel in the second portion, a model to label spectrum maps used for detecting and tracking objects.