Eaton Corporation

Automotive Systems Laboratory, Inc.

Other

Michael Edward Farmer

Michael P Bruce

Craig S Jacobs

James Melvin Slicker

Edward J Gillis

Joseph Spryshak

Galen E Ressler

Christopher Nelson St John

James A Campbell

Shan Cong

John D Prainito

Kevin Jump

Mark L Dell' Eva

Mark L Dell′ Eva

Michael Montag

Daniel M Croft

J Gregory Stanley

Naveed Mahbub

A system or method (collectively 'selection system') is disclosed for selecting attributes for a classifier in a sensor system. The selection system selects attribute types using statistical distributions of the attribute values associated with those attribute types. Attribute types not within the selectively identified subset of attribute types can be filtered out before such data is sent to a classifier. The system can use a test data subsystem for storing and accessing actual sensor data. A distribution analysis subsystem can perform statistical analyses on the test data to identify underlying distributions, and to compare individual attribute types to such distributions. An attribute selection subsystem, wherein said attribute selection subsystem selectively identifies a subset of attribute types from said subset of attribute types.

System or method for selecting classifier attribute types

The present invention relates in general to systems used to process images. In particular, the present invention is an image segmentation system and method used to isolate the segmented image of a target person, animal, or object from an ambient image which includes the target person, animal, or object, in addition to the area surrounding the target. The invention supports the ability of an airbag deployment system to distinguish between different types of occupants by providing such deployment systems with a segmented image of the occupant. The invention is particularly useful at night or in other environments involving inadequate light or undesirable shadows. The invention can use histograms and cumulative distribution functions to perform image thresholding. Morphological erosion and dilation can be used to eliminate optical 'noise' from the image. Gap filling is performed on the basis of the 'momentum' and &#x201c;gravity&#x201d; of regions of similar pixel values. An upper ellipse can generated to represent the upper torso of the occupant. The invention is highly flexible, and can be modified in accordance with the desired use.

Image segmentation system and method

The disclosure describes systems and methods that pertain to interactions between a vehicle and an occupant within the vehicle. More specifically, various systems and methods for enhancing the decisions of automated vehicle applications (collectively 'decision enhancement system') are disclosed. In a safety restraint embodiment, a sensor is used to capture various sensor readings. Sensor readings are typically in the form of images. Occupant information, such as location attributes, motion attributes, and occupant category attributes can be obtained from the sensor readings. Such information can then used by the system to enhance the decisions made by various automated applications.

Decision enhancement system for a vehicle safety restraint application

The present invention relates in general to systems used to determine whether airbag deployment should be disabled or modified due to the proximity of the occupant to the airbag. In particular, the present invention is an image processing system that uses an image and signal process that utilizes real-time streaming video-images from a video camera to determine if the occupant is too close to the air bag, or will be to close the air bag by the time that the airbag deploys. In a crash situation, the system quickly determines whether or not the airbag should be disabled or modified. The process uses a multiple-model Kalman filter to infer three-dimensional information from a sequence of two-dimensional images. The system predicts the position and shape of the occupant at a faster rate than the rate in which the camera collects data.

Image processing system for dynamic suppression of airbags using multiple model likelihoods to infer three dimensional information

The disclosure describes systems and methods that pertain to interactions between a vehicle and an occupant within the vehicle. More specifically, various systems and methods for enhancing the decisions of automated vehicle applications (collectively 'decision enhancement system') are disclosed. In a safety restraint embodiment, a sensor is used to capture various sensor readings. Sensor readings are typically in the form of images. Occupant information, such as location attributes, motion attributes, and occupant category attributes can be obtained from the sensor readings. If the system concludes that the deployment of a safety restraint is potentially justified, an at-risk-zone detector can be used to determine whether or not the occupant will be too close to the deploying safety restraint at the time of deployment for the safety restraint to be safely deployed.

An invention is disclosed to divide a segmented image of an occupant into an upper torso image and a lower torso image. An occupant labeling heuristic can identify each pixel within the segmented image as an upper torso pixel or a lower torso pixel. A k-means module can provide an initial pixel classification by comparing the distance between the particular pixel and an estimated midpoint on the upper torso with the distance between the particular pixel and an estimated midpoint on the lower torso. The iterative parameters estimator can update the mean values for the upper torso and lower torso by performing a conditional likelihood heuristic. Pixels can then be classified as either upper or lower torso pixels by comparing a Mahalonobis distance for each torso. Airbag-related applications can then use the upper torso image to generate occupant characteristics relevant to airbag-related applications.

Occupant labeling for airbag-related applications

The invention can determine whether a collision has occurred, whether an airbag should be deployed, and what the desirable strength of an airbag deployment should be based on occupant characteristics obtained from the segmented image of the occupant. Using interacting multiple model Kalman filters, the invention can infer three-dimensional characteristics from a two-dimensional image, and predict the location, velocity, and acceleration of the occupant. Kalman filters are used with respect to all measurements to incorporate past predictions and measurements into the most recent estimates and predictions in order to eliminate the &ldquo;noise&rdquo; associated with any particular measurement. The system can predict the position and shape of the occupant at a faster rate than the rate at which the sensor collects data. Occupant characteristics can be compared to corresponding predetermined data associated with various occupant modes, including the mode of crashing, to determine if a collision has occurred.

Image processing system for detecting when an airbag should be deployed

The present invention relates in general to systems used to determine whether an airbag should be deployed at full or only partial strength. In particular, the present invention is an image processing system that utilizes real-time streaming video-images from a video camera or other sensor to determine the mass, velocity, and kinetic energy of the occupant at the time that the occupant comes into contact with the deploying airbag. By predicting the kinetic energy of the occupant at the time of impact, an airbag can be deployed at an appropriate strength corresponding to the kinetic energy of the occupant. The kinetic energy of the deploying back at the moment of impact should be equal to the kinetic energy of the occupant. The invention captures the volume of the occupant from an image, and uses volume to calculate the mass of the occupant. A Kalman filter is used with respect to all measurements to incorporate past predictions and measurements into the most recent estimates and predictions in order to eliminate the &ldquo;noise&rdquo; associated with any particular measurement. The system predicts the position and shape of the occupant at a faster rate than the rate at which the camera collects data.

Image processing system for estimating the energy transfer of an occupant into an airbag

A neural network radar processor ( ) comprises a multilayer perceptron neural network ( ) comprising an input layer ( ), a second layer ( ), and at least a third layer ( ), wherein each layer has a plurality of nodes ( ), and respective subsets of nodes ( ) of the second ( ) and third ( ) layers are interconnected so as to form mutually exclusive subnetworks ( ). In-phase and quadrature phase time series from a sampled down-converted FMCW radar signal ( ) are applied to the input layer, and the neural network ( ) is trained so that the nodes of the output layer ( ) are responsive to targets in corresponding range cells, and different subnetworks ( ) are responsive to respectively different non-overlapping sets of target ranges. The neural network is trained with signals that are germane to an FMCW radar, including a wide range of target scenarios as well as leakage signals, DC bias signals, and background clutter signals.

Neural network radar processor

An occupant discrimination system incorporates a position sensor for measuring the distance from a fixed structure within the vehicle to a first surface of an occupant or object on a vehicle seat, and a seat weight sensor for measuring the weight of the occupant or object. The position sensor generates a plurality of signal components that are combined with the weight and vehicle speed related measurements so as to form a signal space, from which a plurality of measures are calculated. A plurality of seat occupancy scenarios are defined and the likelihood of each seat occupancy scenario is functionally related to the plurality of measures. For a given measure state, the likelihood of each of the seat occupancy scenarios is calculated, and the most likely scenario is used to govern the control of a safety restraint system.

Vehicle occupant discrimination system and method

A method and system (,) for detecting vehicle occupant type and position utilizes a single camera unit (,) positioned, for example at the driver or passenger side A-pillar, to generate image data of the front seating area of the vehicle. The present invention distinguishes between objects, forwardly or rearwardly facing child seats, and occupants, by periodically mapping the image taken of the interior of the vehicle into an image profile (,), and utilizing image profile matching with stored profile data (,) to determine the occupant or object type. The system and method of the present invention track occupant type and position in both parallel and perpendicular directions relative to a fixed structure such as the vehicle instrument panel to optimize both the efficiency and safety in controlling deployment of a occupant safety device, such as an air bag (,).

Occupant type and position detection system

A relatively narrow beam of either RF or optical electromagnetic radiation is scanned over a relatively wide azimuthal range. The return signal is processed to detect the range and velocity of each point of reflection. Individual targets are identified by clustering analysis and are tracked in a Cartesian coordinate system using a Kalman filter. The threat to the vehicle for a given target is assessed from estimates of the relative distance, velocity, and size of each target, and one or more vehicular devices are controlled responsive to the assessment of threat so as to enhance the safety of the vehicle occupant. In a preferred embodiment, a quantized linear frequency modulated continuous wave RF signal is transmitted from and received by a multi-beam antenna having an aziumthal range of at least +/-100 degrees and an individual beam width of approximately 10 degrees.

Predictive collision sensing system

An automotive radar incorporates a repetitive randomized equivalent LFM sequence of frequencies for improved immunity to jamming from other automotive radars. Each frequency in the sequence is of sufficient duration to travel round trip over the detection range of the radar. The Doppler shift in the received signal is estimated by performing a spectral analysis on similar frequency components of the received signal, and is then removed from the entire received signal. The received signal is then reordered so as to form an equivalent LFM received signal, and is compared with a similarly reordered image of the transmitted signal so as to estimate the range to the target. The randomization sequence, initial start frequency, or initial start time of the repetitive sequence are varied to minimize the effects of jamming by other radars, and this variation can be directionally dependent.

Vehicle collision radar with randomized FSK waveform

A method and system (10) for detecting vehicle occupant type and position utilizes a single camera unit (12) positioned, for example at the driver or passenger side A-pillar, to generate image data of the front seating area of the vehicle. The present invention distinguishes between objects, forwardly or rearwardly facing child seats, and occupants, by periodically mapping the image taken of the interior of the vehicle into an image profile (104), and utilizing image profile matching with stored profile data (110) to determine the occupant or object type. The system and method of the present invention track occupant type and position in both parallel and perpendicular directions relative to a fixed structure such as the vehicle instrument panel to optimize both the efficiency and safety in controlling deployment of a occupant safety device, such as an air bag (28).

A vehicle collision warning system converts collision threat messages from a predictive collision sensor into intuitive sounds which are perceived by the occupant to be directed from the direction of the potential collision. The type and volume of the sounds are dependent upon the estimated likelihood, severity, and commencement time of the collision. The types of sound are chosen to evoke the proper corrective action by the driver as necessary to avoid the collision or mitigate the effects thereof. Examples of the sounds include a horn, screeching tires, a siren, sounds of various types of vehicles and object, and voice commands. The sounds are stored monaurally and are converted to directional sounds using known techniques.

Growing community of inventors

West Bloomfield, MI, United States of America

Michael Edward Farmer