Abstract: A device for processing data sequences by means of a convolutional neural network is configured to carry out the following steps: receiving an input sequence comprising a plurality of data items captured over time using a sensor, each of said data items comprising a multi-dimensional representation of a scene, generating an output sequence representing the input sequence processed item-wise by the convolutional neural network, wherein generating the output sequence comprises: generating a grid-generation sequence based on a combination of the input sequence and an intermediate grid-generation sequence representing a past portion of the output sequence or the grid-generation sequence, generating a sampling grid on the basis of the grid-generation sequence, generating an intermediate output sequence by sampling from the past portion of the output sequence according to the sampling grid, and generating the output sequence based on a weighted combination of the intermediate output sequence and the input sequence.

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.

Abstract: This disclosure describes systems and techniques for processing radar sensor data. The systems and techniques include acquiring radar sensor data from a radar sensor and processing the radar sensor data by, for example, an artificial neural network to obtain at least one of range radar data or Doppler radar data.

Abstract: A device for processing data sequences by means of a convolutional neural network is configured to carry out the following steps: receiving an input sequence comprising a plurality of data items captured over time using a sensor, each of said data items comprising a multi-dimensional representation of a scene, generating an output sequence representing the input sequence processed item-wise by the convolutional neural network, wherein generating the output sequence comprises: generating a grid-generation sequence based on a combination of the input sequence and an intermediate grid-generation sequence representing a past portion of the output sequence or the grid-generation sequence, generating a sampling grid on the basis of the grid-generation sequence, generating an intermediate output sequence by sampling from the past portion of the output sequence according to the sampling grid, and generating the output sequence based on a weighted combination of the intermediate output sequence and the input sequence.

Abstract: The present disclosure is directed at systems and methods for determining objects around a vehicle. In aspects, a system includes a sensor unit having at least one radar sensor arranged and configured to obtain radar image data of external surroundings to determine objects around a vehicle. The system further includes a processing unit adapted to process the radar image data to generate a top view image of the external surroundings of the vehicle. The top view image is configured to be displayed on a display unit and useful to indicate a relative position of the vehicle with respect to determined objects.

Abstract: A device for processing data sequences by means of a convolutional neural network is configured to carry out the following steps: receiving an input sequence comprising a plurality of data items captured over time using a sensor, each of said data items comprising a multi-dimensional representation of a scene, generating an output sequence representing the input sequence processed item-wise by the convolutional neural network, wherein generating the output sequence comprises: generating a grid-generation sequence based on a combination of the input sequence and an intermediate grid-generation sequence representing a past portion of the output sequence or the grid-generation sequence, generating a sampling grid on the basis of the grid-generation sequence, generating an intermediate output sequence by sampling from the past portion of the output sequence according to the sampling grid, and generating the output sequence based on a weighted combination of the intermediate output sequence and the input sequence.

Abstract: A method is provided for training a machine-learning algorithm which relies on primary data captured by at least one primary sensor. Labels are identified based on auxiliary data provided by at least one auxiliary sensor. A care attribute or a no-care attribute is assigned to each label by determining a perception capability of the primary sensor for the label based on the primary data and based on the auxiliary data. Model predictions for the labels are generated via the machine-learning algorithm. A loss function is defined for the model predictions. Negative contributions to the loss function are permitted for all labels. Positive contributions to the loss function are permitted for labels having a care attribute, while positive contributions to the loss function for labels having a no-care attribute are permitted only if a confidence of the model prediction for the respective label is greater than a threshold.

Abstract: A computer implemented method for determining a location of an object comprises the following steps carried out by computer hardware components: determining a pre-stored map of a vicinity of the object; acquiring sensor data related to the vicinity of the object; determining an actual map based on the acquired sensor data; carrying out image registration based on the pre-stored map and the actual map; carrying out image registration based on the image retrieval; and determining a location of the object based on the image registration.

Abstract: A device for extracting dynamic information comprises a convolutional neural network, wherein the device is configured to receive a sequence of data blocks acquired over time, each of said data blocks comprising a multi-dimensional representation of a scene. The convolutional neural network is configured to receive the sequence as input and to output dynamic information on the scene in response, wherein the convolutional neural network comprises a plurality of modules, and wherein each of said modules is configured to carry out a specific processing task for extracting the dynamic information.

Abstract: A method for validation of an obstacle candidate identified within a sequence of image frames comprises the following steps: A. for a current image frame of the sequence of image frames, determining within the current image frame a region of interest representing the obstacle candidate, dividing the region of interest into sub-regions, and, for each sub-region, determining a Time-To-Contact (TTC) based on at least the current image frame and a preceding or succeeding image frame of the sequence of image frames; B. determining one or more classification features based on the TTCs of the sub-regions determined for the current image frame; and C. classifying the obstacle candidate based on the determined one or more classification features.

Abstract: An image processing method includes: determining a candidate track in an image of a road, wherein the candidate track is modelled as a parameterized line or curve corresponding to a candidate lane marking in the image of a road; dividing the candidate track into a plurality of cells, each cell corresponding to a segment of the candidate track; determining at least one marklet for a plurality of said cells, wherein each marklet of a cell corresponds to a line or curve connecting left and right edges of the candidate lane marking; determining at least one local feature of each of said plurality of cells based on characteristics of said marklets; determining at least one global feature of the candidate track by aggregating the local features of the plurality of cells; and determining if the candidate lane marking represents a lane marking based on the at least one global feature.

Abstract: An image processing method includes: determining a candidate track in an image of a road, wherein the candidate track is modelled as a parameterized line or curve corresponding to a candidate lane marking in the image of a road; dividing the candidate track into a plurality of cells, each cell corresponding to a segment of the candidate track; determining at least one marklet for a plurality of said cells, wherein each marklet of a cell corresponds to a line or curve connecting left and right edges of the candidate lane marking; determining at least one local feature of each of said plurality of cells based on characteristics of said marklets; determining at least one global feature of the candidate track by aggregating the local features of the plurality of cells; and determining if the candidate lane marking represents a lane marking based on the at least one global feature.

Abstract: A method for validation of an obstacle candidate identified within a sequence of image frames comprises the following steps: A. for a current image frame of the sequence of image frames, determining within the current image frame a region of interest representing the obstacle candidate, dividing the region of interest into sub-regions, and, for each sub-region, determining a Time-To-Contact (TTC) based on at least the current image frame and a preceding or succeeding image frame of the sequence of image frames; B. determining one or more classification features based on the TTCs of the sub-regions determined for the current image frame; and C. classifying the obstacle candidate based on the determined one or more classification features.

Abstract: A device for processing data sequences by means of a convolutional neural network is configured to carry out the following steps: receiving an input sequence comprising a plurality of data items captured over time using a sensor, each of said data items comprising a multi-dimensional representation of a scene, generating an output sequence representing the input sequence processed item-wise by the convolutional neural network, wherein generating the output sequence comprises: generating a grid-generation sequence based on a combination of the input sequence and an intermediate grid-generation sequence representing a past portion of the output sequence or the grid-generation sequence, generating a sampling grid on the basis of the grid-generation sequence, generating an intermediate output sequence by sampling from the past portion of the output sequence according to the sampling grid, and generating the output sequence based on a weighted combination of the intermediate output sequence and the input sequence.

Abstract: A device for extracting dynamic information comprises a convolutional neural network, wherein the device is configured to receive a sequence of data blocks acquired over time, each of said data blocks comprising a multi-dimensional representation of a scene. The convolutional neural network is configured to receive the sequence as input and to output dynamic information on the scene in response, wherein the convolutional neural network comprises a plurality of modules, and wherein each of said modules is configured to carry out a specific processing task for extracting the dynamic information.

Abstract: A system for detecting hand gestures in a 3D space comprises a 3D imaging unit. The processing unit generates a foreground map of the at least one 3D image by segmenting foreground from background and a 3D sub-image of the at least one 3D image that includes the image of a hand by scaling a 2D intensity image, a depth map and a foreground map of the at least one 3D image such that the 3D sub-image has a predetermined size and by rotating the 2D intensity image, the depth map and the foreground map of the at least one 3D image such that a principal axis of the hand is aligned to a predetermined axis in the 3D sub-image. Classifying a 3D image comprises distinguishing the hand in the 2D intensity image of the 3D sub-image from other body parts and other objects and/or verifying whether the hand has a configuration from a predetermined configuration catalogue. Further, the processing unit uses a convolutional neural network for the classification of the at least one 3D image.

Abstract: A method for lane detection for a camera-based driver assistance system includes the following steps: image regions in images that are recorded by a camera are identified as detected lane markings if the image regions meet a specified detection criterion. At least two detected lane markings are subjected to a tracking process as lane markings to be tracked. By means of a recursive state estimator, separate progressions are estimated for at least two of the lane markings to be tracked. Furthermore, for each of a plurality of the detected lane markings, a particular offset value is determined, which indicates a transverse offset of the detected lane marking in relation to a reference axis. By means of an additional estimation method, the determined offset values are each associated with one of the separate progressions of the lane markings to be tracked.

Abstract: In a method of generating a training image for teaching of a camera-based object recognition system suitable for use on an automated vehicle which shows an object to be recognized in a natural object environment, the training image is generated as a synthetic image by a combination of a base image taken by a camera and of a template image in that a structural feature is obtained from the base image and is replaced with a structural feature obtained from the template image by means of a shift-map algorithm.

Abstract: A method for lane detection for a camera-based driver assistance system includes the following steps: image regions in images that are recorded by a camera are identified as detected lane markings if the image regions meet a specified detection criterion. At least two detected lane markings are subjected to a tracking process as lane markings to be tracked. By means of a recursive state estimator, separate progressions are estimated for at least two of the lane markings to be tracked. Furthermore, for each of a plurality of the detected lane markings, a particular offset value is determined, which indicates a transverse offset of the detected lane marking in relation to a reference axis. By means of an additional estimation method, the determined offset values are each associated with one of the separate progressions of the lane markings to be tracked.

Abstract: In a method of generating a training image for teaching of a camera-based object recognition system suitable for use on an automated vehicle which shows an object to be recognized in a natural object environment, the training image is generated as a synthetic image by a combination of a base image taken by a camera and of a template image in that a structural feature is obtained from the base image and is replaced with a structural feature obtained from the template image by means of a shift-map algorithm.