# US 2026/0100029 A1 · Method for training a neural network for detecting an object and method for detecting an object via a neural network

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0100029 A1
- Document ID: US20260100029A1
- Status: Published application
- Kind: A1
- Application number: 19/112,280
- Filed: 2023-07-27
- Published: 2026-04-09
- Earliest claimed: 2022-09-15
- Assignees: Sew-Eurodrive GmbH & Co. KG
- Applicants: Sew-Eurodrive GmbH & Co. KG
- Inventors: Thomas SCHÄFER; Leon Rafael SCHÖNFELD
- Web page: https://patentgrep.com/patent/US20260100029A1

## CPC

- G01S 7/417: {involving the use of neural networks}
- G06V 10/82: using neural networks
- G01S 13/867: {Combination of radar systems with cameras}
- G06T 7/70: Determining position or orientation of objects or cameras (camera calibration G06T7/80)
- G06T 2207/10048: Infrared image
- G01S: RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL

## Abstract

In a method for training a neural network for detecting an object, geometric dimensions of a test object from an object class are captured, and during a time period, recordings of the test object are generated by a plurality of cameras. From the captured geometric dimensions and the generated recordings, occupancy maps are generated. By a radar device, a radar signal is transmitted, and a radar signal reflected by the test object is received. The transmitted radar signal and the received radar signal are mixed into a complex baseband to form a mixed signal. A complex four-dimensional mixed spectrum of the mixed signal is calculated. From the complex four-dimensional mixed spectrum, a first complex two-dimensional partial spectrum and a second complex two-dimensional partial spectrum are calculated. The occupancy maps and the partial spectra are fusioned to form training data. The training data are fed to the neural network.

## Claims

### Claim at position 1

1 to 15. (canceled)

### Claim 16

A method for training a neural network for detecting an object, comprising:
capturing geometric dimensions of a test object from an object class;
during a time period, generating recordings of the test object by a plurality of cameras;
generating, from the captured geometric dimensions and the generated recordings, occupancy maps;
transmitting, by a radar device, a radar signal, and receiving, by the radar device, a radar signal reflected by the test object;
mixing the transmitted radar signal and the received radar signal into a complex baseband to form a mixed signal;
calculating a complex four-dimensional mixed spectrum of the mixed signal;
calculating, from the complex four-dimensional mixed spectrum, a first complex two-dimensional partial spectrum and a second complex two-dimensional partial spectrum;
fusioning the occupancy maps and the partial spectra to form training data; and
feeding the training data to the neural network.

### Claim 17 (depends on 16)

The method according to claim 16, wherein the mixed spectrum includes information relating to a distance, an azimuth angle, an elevation angle, and a radial velocity of the test object.

### Claim 18 (depends on 16)

The method according to claim 16, wherein the first partial spectrum includes information relating to a distance and an azimuth angle of the test object, and the second partial spectrum includes information relating to a distance and a radial velocity of the test object.

### Claim 19 (depends on 18)

The method according to claim 18, wherein the first partial spectrum includes a first radar image with information relating to an amount of the distance and the azimuth angle of the test object, the first partial spectrum includes a second radar image with information relating to a phase of the distance and the azimuth angle of the test object, the second partial spectrum includes a third radar image with information relating to an amount of the distance and the radial velocity of the test object, and the second partial spectrum includes a fourth radar image with information relating to a phase of the distance and the radial velocity of the test object.

### Claim 20 (depends on 16)

The method according to claim 16, further comprising moving the test object during the time period.

### Claim 21 (depends on 16)

The method according to claim 16, wherein the occupancy maps are first created in Cartesian coordinates, and the Cartesian coordinates are transformed into polar coordinates.

### Claim 22 (depends on 16)

The method according to claim 16, wherein markings are applied to the test object before the recordings are generated such that the markings are visible in the generated recordings.

### Claim 23 (depends on 16)

The method according to claim 16, wherein the cameras include infrared cameras.

### Claim 24 (depends on 22)

The method according to claim 22, wherein the cameras include infrared cameras, and the markings include infrared markers.

### Claim 25 (depends on 16)

The method according to claim 16, further comprising calculating a respective pose of the test object from the recordings, each pose including a position of the test object and an orientation of the test object, and integrating the calculated poses into the occupancy maps.

### Claim 26 (depends on 16)

The method according to claim 16, further comprising repeating the method for at least one further test object from a further object class, and assigning the occupancy maps and/or the training data to a respective object class.

### Claim 27 (depends on 16)

The method according to claim 16, wherein the neural network includes a convolutional network having an input layer, an output layer, and a plurality of convolutional layers.

### Claim 28 (depends on 16)

A method for detecting an object via a neural network to which training data were previously fed according to the method recited in claim 16, comprising:
transmitting, by a radar sensor, a radar signal, and receiving, by the radar sensor, a radar signal reflected by the object;
mixing the transmitted radar signal and the received radar signal to form a mixed signal;
calculating a mixed spectrum of the mixed signal;
feeding input data including the mixed spectrum to the neural network;
processing the input data in the neural network;
detecting the object and a position of the object by the neural network; and
outputting an object class of the detected object and the detected position of the object by the neural network as output data.

### Claim 29 (depends on 28)

The method according to claim 28, wherein the neural network includes a convolutional network having an input layer, an output layer, and a plurality of convolutional layers, a convolution operation being respectively performed from one layer to the next.

### Claim 30 (depends on 28)

The method according to claim 28, wherein the calculated mixed spectrum includes a distance and an azimuth angle of a radar measurement, first input data including the distance of the radar measurement are fed to the neural network, and second input data including the azimuth angle of the radar measurement are fed to the neural network.

### Claim 31 (depends on 28)

The method according to claim 28, wherein the calculated mixed spectrum includes a distance and a radial velocity of a radar measurement, third input data including the distance of the radar measurement are fed to the neural network, and fourth input data including the radial velocity of the radar measurement are fed to the neural network.

### Claim 32 (depends on 31)

The method according to claim 31, wherein the calculated mixed spectrum includes a distance and a radial velocity of the radar measurement, third input data including the distance of the radar measurement are fed to the neural network, and fourth input data including the radial velocity of the radar measurement are fed to the neural network.
