Method, device, and computer program for detecting an object

Patent metadata

Field	Value
Publication number	US-10824879-B2
Application number	US-201816168217-A
Country	US
Kind code	B2
Filing date	Oct 23, 2018
Priority date	Nov 6, 2017
Publication date	Nov 3, 2020
Grant date	Nov 3, 2020

Abstract

A method for detecting an object with the aid of a system, the system having a plurality of modules connected in series one after the other, at least one connection between a predefinable module and its immediately following module being implemented in each case with the aid of a filter and storage module connected in series, the first module of the modules connected in series being connected to an acquisition unit, and the last module of the modules connected in series being connected to a detection module. The method includes temporal filtering.

First claim

What is claimed is:

1. A method for detecting an object with the aid of a system, the system having a plurality of modules connected in series one after the other, at least one connection between a predefinable module and its immediately following module being implemented with the aid of a filter and storage module connected in series in each case, a first module of the modules connected in series being connected to an acquisition unit, and a last module of the modules connected in series being connected to a detection module, the method comprising:
providing an input variable for the first module of the modules connected in series, with the aid of the acquisition unit; and
successively ascertaining along the modules connected in series, starting with the first module of the modules connected in series, an output variable of each of the modules connected in series as a function of an input variable of the respective module and a parameterization of the respective module, the output variable of the predefinable module being filtered with the aid of the filter module, the filter module ascertaining an output variable as a function of the output variable of the predefinable module and a temporally preceding intermediate variable, the filter module ascertaining a temporally instantaneous intermediate variable as a function of the temporally preceding output variable of the predefinable module and a temporally preceding intermediate variable,
the temporally preceding intermediate variable having been ascertained as a function of a temporally preceding output variable of the predefinable module, with the aid of the filter module, the output variable of the filter module and at least one input variable of one of the preceding modules of the predefinable module being combined into a storage variable with the aid of the storage module and being supplied as an input variable for the immediately following module;
wherein the filter module additionally has a bridging connection, and when filtering the output variable of the predefinable module with the aid of the filter module, the input variable of the filter module is forwarded with the aid of the bridging connection to the output of the filter module and the input variable of the filter module is added at the output of the filter module to the output variable of the filter module.

2. The method as recited in claim 1 , wherein the filter module filters the output variable of the predefinable module as a function of the temporally preceding intermediate variable, the temporally preceding intermediate variable characterizing a temporal development of the output variable of the predefinable module.

3. The method as recited in claim 1 , wherein the filter module includes a normalization module and an activation module and a LSTM module, and the modules connected in series each include a machine learning system, and the temporally instantaneous intermediate variable being ascertained with the aid of the LSTM module as a function of the input variable of the LSTM module and the temporally preceding intermediate variable.

4. The method as recited in claim 1 , wherein the filter module additionally includes a decoding module and the decoding module connects the LSTM module and the output of the filter module, and when the output variable of the predefinable module is filtered with the aid of the filter module, the output variable of the LSTM module is decoded with the aid of the decoding module in such a way that the output variable of the filter module has the same number of dimensions as the input variable of the filter module forwarded with the aid of the bridging connection.

5. The method as recited in claim 1 wherein the connection between the last module of the modules connected in series and the detection module is implemented with the aid of a third storage module, the third storage module combining the output variable of the last module and at least one output variable of a filter module connecting two consecutive modules of the modules connected in series, into one storage variable.

6. The method as recited in claim 1 wherein an object is detected with the aid of a detection module at the output of the last module of the modules connected in series, and an at least partially autonomous machine is controlled as a function of the detected object.

7. A non-transitory machine-readable storage module on which is stored a computer program for detecting an object with the aid of a system, the system having a plurality of modules connected in series one after the other, at least one connection between a predefinable module and its immediately following module being implemented with the aid of a filter and storage module connected in series in each case, a first module of the modules connected in series being connected to an acquisition unit, and a last module of the modules connected in series being connected to a detection module, the computer program, when executed by a processor, causing the processor to perform:
providing an input variable for the first module of the modules connected in series, with the aid of the acquisition unit; and
successively ascertaining along the modules connected in series, starting with the first module of the modules connected in series, an output variable of each of the modules connected in series as a function of an input variable of the respective module and a parameterization of the respective module, the output variable of the predefinable module being filtered with the aid of the filter module, the filter module ascertaining an output variable as a function of the output variable of the predefinable module and a temporally preceding intermediate variable, the filter module ascertaining a temporally instantaneous intermediate variable as a function of the temporally preceding output variable of the predefinable module and a temporally preceding intermediate variable,
the temporally preceding intermediate variable having been ascertained as a function of a temporally preceding output variable of the predefinable module, with the aid of the filter module, the output variable of the filter module and at least one input variable of one of the preceding modules of the predefinable module being combined into a storage variable with the aid of the storage module and being supplied as an input variable for the immediately following module;
wherein the filter module additionally has a bridging connection, and when filtering the output variable of the predefinable module with the aid of the filter module, the input variable of the filter module is forwarded with the aid of the bridging connection to the output of the filter module and the input variable of the filter module is added at the output of the filter module to the output variable of the filter module.

8. A device for detecting an object with the aid of a system, the system having a plurality of modules connected in series one after the other, at least one connection between a predefinable module and its immediately following module being implemented with the aid of a filter and storage module connected in series in each case, a first module of the modules connected in series being connected to an acquisition unit, and a last module of the modules connected in series being connected to a detection module, the device configured to:
provide an input variable for the first module of the modules connected in series, with the aid of the acquisition unit; and
successively ascertain along the modules connected in series, starting with the first module of the modules connected in series, an output variable of each of the modules connected in series as a function of an input variable of the respective module and a parameterization of the respective module,

Assignees

• Bosch Gmbh Robert

Inventors

• Herman Michael
• Wagner Joerg
• Fischer Volker

Classifications

• G06V10/82 Primary

  using neural networks · CPC title

• G06V20/56

  exterior to a vehicle by using sensors mounted on the vehicle · CPC title

• G06F18/24133

  Distances to prototypes · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• G06V10/454

  Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN] · CPC title