Dynamic region of interest identification for vehicles

1 . A method for identifying a region of interest in an environment surrounding a vehicle, the method comprising:

receiving interest data using a first vehicle sensor, wherein the interest data includes a first measurement, and wherein the first vehicle sensor includes at least one of: a radar sensor and a light detection and ranging (LIDAR) sensor;

determining an intended path of the vehicle;

identifying the region of interest in the environment surrounding the vehicle based at least in part on at least one of: the interest data and the intended path of the vehicle, wherein identifying the region of interest in the environment surrounding the vehicle further comprises:

identifying an object of interest in the environment surrounding the vehicle based at least in part on the first measurement;

determining a predicted path of the object of interest based at least in part on the first measurement;

calculating a time-to-collision of the object of interest with the vehicle based at least in part on the predicted path of the object of interest;

determining a probability of collision based at least in part on the predicted path of the object of interest, an uncertainty of the predicted path of the object of interest, and the time-to-collision of the object of interest with the vehicle; and

identifying the region of interest to include the object of interest in response to determining that the probability of collision is greater than or equal to a predetermined collision probability threshold;

performing a second measurement of the region of interest using a second vehicle sensor, wherein performing the second measurement of the region of interest using the second vehicle sensor further comprises:

capturing a first image of the environment surrounding the vehicle using the second vehicle sensor, wherein the first image has a first image resolution, wherein the first image includes at least the region of interest, and wherein the second vehicle sensor is a camera;

determining a priority level of the region of interest, wherein the priority level includes at least one of: a high priority level and a low priority level;

caching the first image in a non-transitory memory in response to determining that the priority level is the low priority level; and

generating a second image of the environment surrounding the vehicle in response to determining that the priority level is the high priority level, wherein generating the second image further comprises:

comparing the first image resolution to a maximum image resolution of the camera;

generating an upscaled image of the environment surrounding the vehicle in response to determining that the first image resolution is equal to the maximum image resolution of the camera, wherein the upscaled image includes the region of interest, and wherein the upscaled image of the environment is upscaled using a machine learning super resolution algorithm; and

capturing a high-resolution image of the environment surrounding the vehicle in response to determining that the first image resolution is less than the maximum image resolution of the camera, wherein the high-resolution image includes the region of interest, wherein the high-resolution image of the environment has a second image resolution, and wherein the second image resolution is greater than the first image resolution; and

performing a data processing task based at least in part on the second measurement, wherein the data processing task includes detection of a lane line on a roadway; and

controlling a motion of the vehicle using a driver assistance feature based at least in part on the lane line.

2 . The method of claim 1 , wherein identifying the region of interest in the environment surrounding the vehicle further comprises:

identifying an object of interest in the environment surrounding the vehicle based at least in part on the first measurement;

determining a predicted path of the object of interest based at least in part on the first measurement;

calculating a time-to-collision of the object of interest with the vehicle based at least in part on the predicted path of the object of interest;

determining a probability of collision based at least in part on the predicted path of the object of interest, an uncertainty of the predicted path of the object of interest, and the time-to-collision of the object of interest with the vehicle; and

identifying the region of interest to include the object of interest in response to determining that the probability of collision is greater than or equal to a predetermined collision probability threshold.

3 . The method of claim 1 , wherein determining the intended path of the vehicle further comprises:

receiving one or more occupant inputs from an occupant of the vehicle using one or more vehicle input devices;

performing one or more vehicle dynamics measurements with one or more vehicle dynamics sensors;

determining a location of the vehicle using a global navigation satellite system (GNSS); and

determining the intended path of the vehicle based at least in part on at least one of: the one or more occupant inputs, the one or more vehicle dynamics measurements, and the location of the vehicle.

4 . The method of claim 3 , wherein identifying the region of interest in the environment surrounding the vehicle further comprises:

identifying the region of interest in the environment surrounding the vehicle, wherein the region of interest includes at least a portion of the intended path of the vehicle.

5 . The method of claim 1 , wherein receiving interest data further comprises:

receiving one or more region cues from a remote server system, wherein the first measurement is the one or more region cues.

6 . The method of claim 5 , wherein identifying the region of interest in the environment surrounding the vehicle further comprises:

identifying the region of interest in the environment surrounding the vehicle based at least in part on the one or more region cues, wherein the one or more region cues includes a location of a missing map element from the remote server system.

7 . The method of claim 5 , wherein identifying the region of interest in the environment surrounding the vehicle further comprises:

identifying the region of interest in the environment surrounding the vehicle based at least in part on the one or more region cues, wherein the one or more region cues includes at least one crowdsourced region of interest parameter, wherein the at least one crowdsourced region of interest parameter is determined by the remote server system using crowdsourcing.

8 . A system for identifying a region of interest in an environment surrounding a vehicle, the system comprising:

a first vehicle sensor;

a second vehicle sensor, wherein the second vehicle sensor includes a camera;

a vehicle graphics processing unit (GPU); and

a vehicle controller in electrical communication with the first vehicle sensor, the second vehicle sensor, and the vehicle GPU, wherein the vehicle controller is programmed to:

receive interest data using the first vehicle sensor;

identify the region of interest in the environment surrounding the vehicle based at least in part on the interest data; and

perform a second measurement of the region of interest using the second vehicle sensor, wherein to perform the second measurement of the region of interest using the second vehicle sensor, the vehicle controller is further programmed to:

capture a first image of the environment surrounding the vehicle using the camera, wherein the first image has a first image resolution, wherein the first image includes at least the region of interest;

determine a priority level of the region of interest, wherein the priority level includes at least one of: a high priority level and a low priority level;

cache the first image in a non-transitory memory of the vehicle controller in response to determining that the priority level is the low priority level; and

generate a second image of the environment surrounding the vehicle in response to determining that the priority level is the high priority level, wherein to generate the second image of the environment surrounding the vehicle, the vehicle controller is further programmed to:

compare the first image resolution to a maximum image resolution of the camera;

generate an upscaled image of the environment surrounding the vehicle in response to determining that the first image resolution is equal to the maximum image resolution of the camera, wherein the upscaled image includes the region of interest, and wherein the upscaled image of the environment is upscaled using a machine learning super resolution algorithm executed using the vehicle GPU; and

capture a high-resolution image of the environment surrounding the vehicle in response to determining that the first image resolution is less than the maximum image resolution of the camera, wherein the high-resolution image includes the region of interest, wherein the high-resolution image of the environment has a second image resolution, and wherein the second image resolution is greater than the first image resolution;

perform a data processing task based at least in part on the second measurement, wherein the data processing task includes detection of a road geometry; and

control a motion of the vehicle using an automated driving feature based at least in part on the road geometry.

9 . The system of claim 8 , wherein the first vehicle sensor includes a perception sensor, and wherein to receive interest data using the first vehicle sensor, the vehicle controller is further programmed to:

perform a first measurement using the perception sensor, wherein the interest data is the first measurement.

10 . The system of claim 9 , wherein to identify the region of interest in the environment surrounding the vehicle, the vehicle controller is further programmed to:

identify an object of interest in the environment surrounding the vehicle based at least in part on the first measurement;

determine a predicted path of the object of interest based at least in part on the first measurement;

calculate a time-to-collision of the object of interest with the vehicle based at least in part on the predicted path of the object of interest;

determine a probability of collision based at least in part on an uncertainty of the predicted path of the object of interest and the time-to-collision of the object of interest with the vehicle; and

identify the region of interest to include the object of interest in response to determining that the probability of collision is greater than or equal to a predetermined collision probability threshold.

11 . The system of claim 8 , wherein the first vehicle sensor is a vehicle communication system, and wherein to receive interest data using the first vehicle sensor, the vehicle controller is further programmed to:

receive one or more region cues from a remote server system using the vehicle communication system, wherein the interest data is the one or more region cues; and

transmit at least one vehicle region of interest parameter to the remote server system using the vehicle communication system for crowdsourcing by the remote server system.

12 . The system of claim 11 , wherein to identify the region of interest, the vehicle controller is further programmed to:

identify the region of interest in the environment surrounding the vehicle based at least in part on the one or more region cues, wherein the one or more region cues includes a location of a missing map element from the remote server system; and

identify the region of interest in the environment surrounding the vehicle based at least in part on the one or more region cues, wherein the one or more region cues includes at least one crowdsourced region of interest parameter, wherein the at least one crowdsourced region of interest parameter is determined by the remote server system using crowdsourcing.

13 . A method for identifying a region of interest in an environment surrounding a vehicle, the method comprising:

performing a first measurement of the environment surrounding the vehicle using a first vehicle sensor, wherein the first vehicle sensor is a perception sensor including at least one of: a radar sensor and a light detection and ranging (LIDAR) sensor;

identifying the region of interest in the environment surrounding the vehicle based at least in part on the first measurement, wherein identifying the region of interest in the environment surrounding the vehicle further comprises:

identifying an object of interest in the environment surrounding the vehicle based at least in part on the first measurement;

determining a predicted path of the object of interest based at least in part on the first measurement;

calculating a time-to-collision of the object of interest with the vehicle based at least in part on the predicted path of the object of interest;

determining a probability of collision based at least in part on an uncertainty of the predicted path of the object of interest and the time-to-collision of the object of interest with the vehicle; and

identifying the region of interest to include the object of interest in response to determining that the probability of collision is greater than or equal to a predetermined collision probability threshold;

performing a second measurement of the region of interest using a second vehicle sensor, wherein the second vehicle sensor is a camera, wherein performing the second measurement of the region of interest using the second vehicle sensor further comprises:

capturing a first image of the environment surrounding the vehicle using the camera, wherein the first image has a first image resolution, wherein the first image includes at least the region of interest;

comparing the first image resolution to a maximum image resolution of the camera;

generating an upscaled image of the environment surrounding the vehicle in response to determining that the first image resolution is equal to the maximum image resolution of the camera, wherein the upscaled image includes the region of interest, and wherein the upscaled image of the environment is upscaled using a machine learning super resolution algorithm; and

capturing a high-resolution image of the environment surrounding the vehicle in response to determining that the first image resolution is less than the maximum image resolution of the camera, wherein the high-resolution image includes the region of interest, wherein the high-resolution image of the environment has a second image resolution, and wherein the second image resolution is greater than the first image resolution;

performing a data processing task based at least in part on the second measurement, wherein the data processing task includes detection of a lane line on a roadway; and

controlling a motion of the vehicle using a driver assistance feature based at least in part on the lane line.