公开(公告)号：US20230236030A1

公开(公告)日：2023-07-27

申请号：US18157446

申请日：2023-01-20

Applicant: Aptiv Technologies Limited

Inventor： Christian Nunn ,  Dennis Mueller ,  Lutz Roese-Koerner ,  Pascal Colling

IPC: G01C21/34 ,  G06N3/0464

CPC classification number: G01C21/3476 ,  G01C21/3461 ,  G01C21/3423 ,  G06N3/0464

Abstract: Provided is a computer-implemented method of determining a point of interest and/or a road type in a map, comprising the steps of: acquiring processed sensor data collected from one or more vehicles; extracting from the processed sensor data a set of classification parameters; and determining based on the set of classification parameters one or more points of interest (POI) and its geographic location and/or one or more road types.

公开(公告)号：US20230101472A1

公开(公告)日：2023-03-30

申请号：US17934897

申请日：2022-09-23

Applicant: Aptiv Technologies Limited

Inventor： Pascal Colling ,  Peet Cremer

IPC: B60W30/12 ,  B60W40/06

Abstract: This disclosure describes methods and techniques for estimating lanes for a vehicle. The methods and techniques include determining a first preliminary estimate of lanes based on a plurality of lane markings at a location of the vehicle, determining a second preliminary estimate of lanes based on a plurality of trails of objects at the location of the vehicle, comparing the first preliminary estimate of lanes and the second preliminary estimate of lanes, and determining a final estimate of lanes at the location of the vehicle based on the comparing.

公开(公告)号：US20220245955A1

公开(公告)日：2022-08-04

申请号：US17579536

申请日：2022-01-19

Applicant: Aptiv Technologies Limited

Inventor： Ido Freeman ,  Pascal Colling

IPC: G06V20/70 ,  G06V10/82 ,  G06T7/12 ,  G06T7/13 ,  G06T7/62 ,  G06V10/774 ,  G06V10/764 ,  G06V10/26

Abstract: A method is provided for classifying pixels of an image. An image comprising a plurality of pixels is captured by a sensor device. A neural network is used for estimating probability values for each pixel, each probability value indicating the probability for the respective pixel being associated with one of a plurality of predetermined classes. One of the classes is assigned to each pixel of the image based on the respective probability values to create a predicted segmentation map. For training the neural network, a loss function is generated by relating the predicted segmentation map to ground truth labels. Furthermore, an edge detection algorithm is applied to at least one of the predicted segmentation maps and the ground truth labels, wherein the edge detection algorithm predicts boundaries between objects. Generating the loss function is based on a result of the edge detection algorithm.