Siemens Aktiengesellschaft

Conti Temic Microelectronic Gmbh

Continental Automotive Gmbh

Siemens Corporation

Continental Automotive France

Siemesn Corporate Research, Inc.

Gregory Baratoff

Thorsten Köhler

Ying Zhu

Martin Pellkofer

Frederik Lange

Christian Exner

Alexander Bachmann

Robert Stueck

Dorin Comaniciu

Maneesh Kumar Singh

Martin Randler

Kazunori Okada

Benedicte Bascle

Michael Feser

Ludwig Ertl

Claus Bahlmann

Thomas Brandmeier

Thorsten Koehler

Jan Giebel

Alain Giralt

Sophie Rony

Andreas Raith

Rudolf Ertlmeier

Andre Rieder

David Jean Pascal Acunzo

Laure Bajard

Jochen Schmidt

In a method for determining a depth map from stereo images, the disparity for a pixel is selected from a predefined quantity or set of predefined discrete disparity values that are distributed over the entire predefined disparity value range, whereby the distribution is non-uniform or has at least two different distances or intervals between different adjacent disparity values. This method makes it possible to more precisely determine (with finer resolution) especially only those disparities for which a more precise determination is required.

Method for determining depth maps from stereo images with improved depth resolution in a range

In a method for determining a depth map from stereo images, the disparity for a pixel is determined from a discrete quantity of predefined disparity values that are distributed over the entire predefined disparity value range. The distribution has at least two different distances between different pairs of adjacent disparity values. In this method, it is possible that only those disparities for which a more precise determination is required, will be more precisely calculated.

A device for redundant object detection includes a stereo camera having two image sensors, a first processor for detecting objects by stereo image evaluation of image data from the two image sensors, and a second processor for classifying the detected objects by evaluation of image data from at least one of the two image sensors. A method and a program for such object detection and classification are also provided.

Redundant object detection for driver assistance systems

To improve the performance for determining obstruction of a stereoscopic system using two cameras or more, a hybrid of local and semi-global methods is provided. For each stereoscopic image formed from simultaneous left and right images, a breakdown of each left and right image into corresponding sectors is applied. Obstruction level is determined by a disparity map by sector, based on left or right images, and in which a disparity is assigned to each pixel corresponding to the best matching score. A determination of density by sector of the disparity map is carried out by reference to a fraction of pixels with a disparity considered to be valid. A state of obstruction of at least one camera is determined based on a weighted average of the probabilities of obscuring of the sectors of the disparity map obtained by comparison between the density of the sectors and a predefined density level.

Method for determining a state of obstruction of at least one camera installed in a stereoscopic system

A method and a device for detecting an imminent rollover of a vehicle. The method comprises taking a series of images of the surroundings of the vehicle, in particular of the road ahead, with a camera; determining and tracking point features in the series of images; eliminating point features which are located on objects that move relative to the stationary surroundings of the vehicle; and computing at least one parameter of the rotational movement of the camera (roll, pitch and/or yaw rate) taking into account the remaining point features.

Method for detecting an imminent rollover of a vehicle

A speed limit assistant (SLA) system includes a camera based SLA, a map based SLA, and a fusion unit. The camera based SLA is configured to determine a first set of probabilities for an input image, wherein the probabilities indicate how likely the image includes a discrete set of speed limit signs. The map based SLA is configured to determine a second set of probabilities for an input coordinate, wherein the probabilities indicate how likely the coordinate is to correspond to one of a discrete set of speed limits. The fusion unit is configured to perform a Bayesian fusion on the first and second set of probabilities to determine a final speed limit from the discrete set of speed limits.

Multi-modal speed limit assistant

A method for vehicle detection and tracking includes acquiring video data including a plurality of frames, comparing a first frame of the acquired video data against a set of one or more vehicle detectors to form vehicle hypotheses, pruning and verifying the vehicle hypotheses using a set of course-to-fine constraints to detect a vehicle, and tracking the detected vehicle within one or more subsequent frames of the acquired video data by fusing shape template matching with one or more vehicle detectors.

System and method for vehicle detection and tracking

An assistance system for motor vehicles has at least one monitoring unit monitoring a space external to a vehicle, in particular in a direction of travel of the motor vehicle, with a warning zone being defined in the monitored external space on each side of a lane. The warning zone is delimited by an earliest warning line and a latest warning line encompassing the respective lane marking or, as the case may be, boundary. An evaluation unit receives the data transmitted by the monitoring unit and generates a warning signal for an acoustic, visual and/or haptic warning and/or a steering intervention if the evaluation unit detects that a defined warning threshold lying within the warning zone is exceeded. The warning signal is issued only as a function of further planned driver-adaptive conditions.

Assistance system for motor vehicles

Adaptive vision-based vehicle detection methods, taking into account the lighting context of the images are disclosed. The methods categorize the scenes according to their lighting conditions and switch between specialized classifiers for different scene contexts. Four categories of lighting conditions have been identified using a clustering algorithm in the space of image histograms: Daylight, Low Light, Night, and Saturation. Trained classifiers are used for both Daylight and Low Light categories, and a tail-light detector is used for the Night category. Improved detection performance by using the provided context-adaptive methods is demonstrated. A night time detector is also disclosed.

Context adaptive approach in vehicle detection under various visibility conditions

Methods for image processing for detecting and recognizing an image object include detecting an image object using pose-specific object detectors, and performing fusion of the outputs from the pose-specific object detectors. The image object is recognized using pose-specific object recognizers that use outputs from the pose-specific object detectors and the fused output of the pose-specific object detectors; and by performing fusion of the outputs of the pose-specific object recognizers to recognize the image object.

Systems and methods for face detection and recognition using infrared imaging

Growing community of inventors

Wangen, Germany

Gregory Baratoff