Aptiv Technologies Ag

Alessandro Cennamo

Florian Kaestner

Mirko Meuter

Stephanie Lessmann

Yu Su

Weimeng Zhu

Jan Siegemund

Adrian Becker

Markus Schoeler

Arne Grumpe

Florian Kästner

Sven Labusch

Saeid Khalili Dehkordi

Zhiheng Niu

Simon Roesler

Moritz Luszek

Dominic Spata

Jittu Kurian

Marco Braun

Kevin Kollek

The present disclosure provides systems and techniques directed at object detection in a vehicle. In aspects, techniques include capturing current radar image data. The current radar image data includes at least one current point cloud. The current point cloud includes at least one current object point being related to an object, and each current object point includes spatial information related to the object. The techniques further include retrieving previous radar image data. The previous radar image data includes at least one previous point cloud. The previous point cloud includes at least one previous object point being related to the object, and each previous object point includes spatial information related to the object. The techniques further include concatenating the information from the current radar image data and the information from the previous radar image data to derive enhanced radar image data using a recurrent neural network.

Object detection in a vehicle

A computer implemented method for detection of objects in a vicinity of a vehicle comprises the following steps carried out by computer hardware components: acquiring radar data from a radar sensor; determining a plurality of features based on the radar data; providing the plurality of features to a single detection head; and determining a plurality of properties of an object based on an output of the single detection head.

Methods and systems for detection of objects in a vicinity of a vehicle

A method is provided for detecting objects by using a radar sensor and by using a device configured to establish a neural network. A plurality of raw radar data points is captured. At least one object class is defined comprising a predefined object type and a geometrical shape for enclosing an object. Via a first stage of the neural network, a semantic segmentation is performed for the data points regarding the object class and background, and for each data point, a rough approximation is estimated for a spatial condition of the geometrical shape. Based on this rough approximation and via a second stage of the neural network, a subset of the data points is selected based on the semantic segmentation, and for each data point of the subset, a refined approximation for the spatial condition of the geometrical shape and a confidence score for the refined approximation are estimated.

Method and device for detecting objects

Growing community of inventors

Wuppertal, Germany

Alessandro Cennamo