US 12,386,061 B2
Apparatus and method for identifying objects in radar signals
Franz-Xaver Hofele, Donzdorf (DE)
Assigned to HENSOLDT SENSORS GmbH, Taufkirchen (DE)
Filed by HENSOLDT SENSORS GMBH, Taufkirchen (DE)
Filed on Jan. 13, 2022, as Appl. No. 17/647,860.
Claims priority of application No. 21151319 (EP), filed on Jan. 13, 2021; application No. 21151334 (EP), filed on Jan. 13, 2021; application No. 21151340 (EP), filed on Jan. 13, 2021; application No. 21151343 (EP), filed on Jan. 13, 2021; and application No. 21151350 (EP), filed on Jan. 13, 2021.
Prior Publication US 2022/0244376 A1, Aug. 4, 2022
Int. Cl. G01S 13/88 (2006.01); G01S 7/41 (2006.01); G01S 13/50 (2006.01); G06N 20/20 (2019.01)
CPC G01S 13/88 (2013.01) 5 Claims
OG exemplary drawing
 
1. An apparatus for automatic identification of a bird in radar signals, the apparatus comprising:
a processor executing a visualization module (110) and a classification module (120),
the visualization module (110) being configured to receive spectral data of the radar signals, the spectral data including a spectrum (spe) as function of frequencies and a cepstrum (cep) as function of quefrencies, and configured to visualize structures in the spectrum (spe) corresponding to bird-features by
generating a first modified spectrum (spe1) by keeping a predefined number (number_50) of data points (V1) with largest amplitude values in the spectrum (spe), while setting all other amplitude values to a default value,
generating a second modified spectrum (spe1a, spe1b, spe2) by repeated averaging of consecutive amplitude values in the first modified spectrum (spe1), and
defining a threshold (T) based on a statistical upper limit for variations of the spectrum (spe) in a range (dS) spaced from a maximal value (Vmax); and
the classification module (120) being configured to identify the bird based on at least in part on a result produced by the visualization module (110), by testing for presence of more than 35 bird-features, wherein the testing for the presence of the more than 35 bird-features comprises determining
a number_6 of data points on one or both sides of a maximum in the first modified spectrum (spe1) that are all greater or equal to the threshold (T) or all greater than the default value,
a number_7 of data points at least one spectral edge in the first modified spectrum (spe1) that are all greater or equal to the threshold (T) or all greater than the default value,
a number_15 given by a sum of differences, wherein, when the spectrum (spe) is split into multiple blocks, each difference is a difference between a maximum and a minimum of amplitude values within a respective block,
a number_5 of crossings of the threshold (T) in the spectrum (spe),
a number_16 based on a standard deviation of data points of the spectrum (spe) while omitting data points that, starting from maximal value (Vmax), are larger than the threshold (T) in an uninterrupted order,
entries (sort_30, sort_50) at one or more predetermined positions (30th, 50th) in an ordered statistic, wherein the data points of the spectrum (spe) are ordered in size,
the number_16 is at least 2.2 and at most 4.1, which defines a first bird-feature,
the number 15 is at least 63 and at most 95, which defines a second bird-feature,
the number_5 is at least zero and at most 6, which defines a third bird-feature,
a number_4 is at least 1 and at most 9, wherein the number 4 is the number of spectral values in the range (dS) used to define the threshold (T), which defines a fourth bird-feature,
wherein the classification module (120) is further configured to identify the bird if the following drone-features is not satisfied
the number 5 is at least 10 or the number 15 is at least 110, which defines a first drone-feature,
wherein, for a number of samples that differ from 256 samples, parameters are adjusted accordingly.