Systems and methods for detecting traffic objects

Patent metadata

Field	Value
Publication number	US-12293589-B2
Application number	US-202217817495-A
Country	US
Kind code	B2
Filing date	Aug 4, 2022
Priority date	Aug 4, 2022
Publication date	May 6, 2025
Grant date	May 6, 2025

Abstract

Systems and methods for controlling a vehicle. The systems and methods receive image data from at least one camera of the vehicle, detect a 2D measurement location of a static object in an image plane of the camera using the image data, receive an input vector of measurements from a sensor system of the vehicle, predict a predicted 3D location of the static object using a Unscented Kalman Filter (UKF) that incorporates a motion model for the vehicle and further using the 2D measurement location of the static object, and the input vector, and control at least one vehicle feature based on the predicted 3D location of the static object.

First claim

What is claimed is:

1. A method of controlling a vehicle, the method comprising:
receiving, via at least one processor, image data from at least one camera of the vehicle;
detecting, via the at least one processor, a 2D measurement location of a static object in an image plane of the camera using the image data;
receiving, via the at least one processor, an input vector determined in response to a steering angle detected by a steering angle sensor of the vehicle, a wheel speed detected by a wheel speed sensor of the vehicle and an acceleration measured by an inertial measurement unit of the vehicle;
predicting, via the at least one processor, a predicted 3D location of the static object using an Unscented Kalman Filter (UKF) that incorporates a motion model for the vehicle and further using the 2D measurement location of the static object, and the input vector wherein the predicting the 3D location of the static object using the UKF includes a prediction step and an update step, wherein the prediction step performs the following:
constructing a prediction sigma point matrix,
propagating the prediction sigma point matrix through the motion model to obtain propagated sigma points,
determining an estimated 3D location of the static object using the propagated sigma points,
estimating a propagated estimation error covariance using the propagated sigma point; and
wherein the update step performs the following:
constructing an update sigma point matrix using the estimated 3D location of the static object and the propagated estimation error covariance,
propagating the update sigma point matrix through a measurement model to obtain a measurement sigma point matrix representing predicted 2D measurement locations of the static object in the image plane of the camera,
determining an updated 3D location of the static object using a disparity between the measurement sigma point matrix and the 2D measurement location of the static object in the image plane of the camera;
controlling, via the at least one processor, at least one vehicle feature based on the predicted 3D location of the static object, wherein the UKF is initialized based on a set of sigma points that are generated using a first detection of the 2D measurement location of the static object in the image plane and a range prior, and
wherein the set of sigma points are used to generate a prior 3D location for an initialization of the UKF, and wherein a disparity between the first detection of the 2D location measurement and a predicted 2D state measurement is used to update the estimated 3D location of the static object such that an updated 3D state and an updated covariance are provided and wherein a subsequent iteration of the UKF is performed in response to the updated 3D state and the updated covariance.

2. The method of claim 1 , wherein the range prior includes an average range value for a first 3D location of the static object relative to the vehicle.

3. The method of claim 1 , wherein the range prior is determined based on an average range in which the static object is initially detected, an estimate of an average range of the static object based on a size of a bounding box provided from the detecting of the 2D measurement location of the static object, an association with other tracked static objects of the same kind and an area of the 2D measurement location of the static object, or a detection of the static object using 3D perception modalities.

4. The method of claim 1 , comprising recursively updating the predicted 3D location of the static object using new detections of the 2D measurement location of the static object in the image plane and a prior prediction of the 3D location of the static object using the UKF.

5. The method of claim 1 , wherein the static object is a Traffic Control Device (TCD).

6. The method of claim 1 , wherein a path of the vehicle is determined in response to the predicted 3D location of the static object and wherein the at least one vehicle feature includes a steering control, a propulsion control and a braking control.

7. The method of claim 1 , comprising assigning a lane to a TCD and controlling the at least one vehicle feature is responsive to a state of the TCD when the TCD is assigned to a same lane as that of the vehicle.

8. The method of claim 1 , wherein the input vector includes a value of angular velocity of the vehicle and a value of linear velocity of the vehicle.

9. The method of claim 1 , wherein the UKF uses the motion model and a measurement model to predict a predicted 2D location of the static object in the image plane and predicts the predicted 3D location of the static object based on a disparity between the predicted 2D location of the static object and the measurement 2D location of the static object.

10. A system for controlling a vehicle, the system comprising:
at least one camera;
a sensor system;
at least one processor in operable communication with the sensor system and the at least one camera, wherein the at least one processor is configured to execute program instructions, wherein the program instructions are configured to cause the at least one processor to:
receive image data from the at least one camera of the vehicle;
detect a 2D measurement location of a static object in an image plane of the camera using the image data;
receive an input vector of measurements from the sensor system of the vehicle wherein the input vector is determined in response to a steering angle detected by a steering angle sensor of the vehicle, a wheel speed detected by a wheel speed sensor of the vehicle and an acceleration measured by an inertial measurement unit;
predict a predicted 3D location of the static object using a Unscented Kalman Filter (UKF) that incorporates a motion model for the vehicle and further using the 2D measurement location of the static object, and the input vector wherein the program instructions are configured to cause the at least one processor to predict the 3D location of the static object using the UKF using a prediction step and an update step, wherein the prediction step performs the following:
constructing a prediction sigma point matrix,
propagating the prediction sigma point matrix through the motion model to obtain propagated sigma points,
determining an estimated 3D location of the static object using the propagated sigma points,
estimating a propagated estimation error covariance using the propagated sigma point:
constructing an update sigma point matrix using the estimated 3D location of the static object and the propagated estimation error covariance,
propagating the update sigma point matrix through a measurement model to obtain a measurement sigma point matrix representing predicted 2D measurement locations of the static object in the image plane of the camera,
determining an updated 3D location of the static object using a disparity between the measurement sigma point matrix and the 2D measurement location of the static object in the image plane of the camera;
control at least one vehicle feature based on the predicted 3D location of the static object, initialize the UKF based on a set of sigma points that are generated using a first detection of the 2D measurement location of the static object in the image plane and a range prior, and

wherein the set of sigma points are used to generate a prior 3D location for an initialization of the UKF, and wherein a disparity between the first detection of the 2D location measurement and a predicted 2D state measurement is used to update the estimated 3D location of the static object such that an updated 3D state and an updated covariance are provided and wherein a subsequent iteration of the UKF is perform

Assignees

• Gm Global Tech Operations Llc

Inventors

• Menaker Yonattan
• Ayzik Sharon
• Amit Ido
• Fomin Max

Classifications

• B60W2420/403

  Image sensing, e.g. optical camera · CPC title

• B60W40/105

  Speed · CPC title

• B60W2554/4049

  Relationship among other objects, e.g. converging dynamic objects · CPC title

• G06V20/588

  Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road · CPC title

• G06V20/56

  exterior to a vehicle by using sensors mounted on the vehicle · CPC title