Systems for calibrating vehicle-based image capture devices
Systems, devices, and methods for calibration of image capture devices at vehicles are described. A calibration polygon is determined based on key points in image data. The calibration polygon can be regularly or continuously calibrated in an ongoing matter based on additional key points in later captured image data. The calibration polygon can be monitored continuously or periodically, to determine when recalibration is required. Calibration for road-facing image capture devices and cabin-facing capture devices is described.
1 . A system, comprising:
at least one image capture device positioned at a vehicle and oriented to face an interior of the vehicle;
at least one processor; and
at least one non-transitory processor-readable storage medium storing processor-executable instructions which when executed by the at least one processor cause the system to:
perform a calibration process which includes:
access first telematics data for the vehicle;
capturing a first image data series by the at least one image capture device, the first image data series representing an interior of the vehicle, the first image data series and the first telematics data representing a first period of time;
determining, by the at least one processor, whether suitability criteria are satisfied for at least the first telematics data;
when the suitability criteria are not satisfied, restarting the calibration process; and
when the suitability criteria are satisfied:
for each image in the first image data series, applying at least one detection model to identify at least one bounding box around at least one human represented in the first image data series;
cluster bounding boxes of each human of the at least one human across the first image data series to identify a respective bounding box cluster for each human of the at least one human;
aggregate a position of each bounding box in each bounding box cluster to determine an aggregated position of each human of the at least one human; and
select one human of the at least one human, the select human positioned at a driver's seat of the vehicle; and
determine a calibration polygon focused on the aggregated position of the select human positioned at a driver's seat of the vehicle; and
perform a monitoring process which includes:
accessing second telematics data for the vehicle;
capturing second image data by the at least one image capture device, the second image data representing the interior of the vehicle, the second image data and the second telematics data representing a common time after the first period of time;
determining, by the at least one processor, whether the suitability criteria are satisfied for at least the second telematics data;
when the suitability criteria are not satisfied, restarting the monitoring process; and
when the suitability criteria are satisfied:
applying the at least one detection model to identify a position of the select human represented in the second image data;
compare the position of the select human in the second image data to the aggregated position of the select human;
when the position of the select human in the second image data deviates from the aggregated position of the select human in excess of a deviation threshold, perform the calibration process again; and
when the position of the select human in the second image data does not deviate from the aggregated position of the select human in excess of a deviation threshold, maintain the calibration polygon focused on the position of the select human.
2 . The system of claim 1 , wherein the first telematics data includes location data over time or speed data indicative of movement speed of the vehicle for the first period of time, and the suitability criteria includes the movement speed of the vehicle exceeding a speed threshold.
3 . The system of claim 1 , wherein the first telematics data includes location data indicative of a location of the vehicle, and the suitability criteria includes the location of the vehicle matching a location classification.
4 . The system of claim 1 , wherein determining whether suitability criteria are satisfied for at least the first telematics data comprises determining whether the suitability criteria are satisfied for the first telematics data and the first image data series.
5 . The system of claim 4 , wherein determining whether the suitability criteria are satisfied for the first image data series comprises determining whether a driver of the vehicle is in a driving posture.
6 . The system of claim 1 , wherein the processor-executable instructions further cause the at least one processor to: determine a region of interest within a field of view of the at least one image capture device, based on the calibration polygon.
7 . The system of claim 6 , wherein the processor-executable instructions further cause the at least one processor to crop image data from the at least one image capture device to the region of interest.
8 . The system of claim 6 , wherein the processor executable instructions further cause the at least one processor to execute a trained detection model on the region of interest for image data from the at least one image capture device.
9 . The system of claim 1 , further comprising at least one actuator coupled to the at least one image capture device, wherein the processor-executable instructions further cause the system to orient, by the at least one actuator, the at least one image capture device to align a field of view of the at least one image capture device based on the calibration polygon.
10 . The system of claim 1 , further comprising a user interface to output, to a user, the calibration polygon.
11 . The system of claim 1 , wherein the deviation threshold comprises a pixel displacement threshold, and the monitoring polygon is within the deviation threshold if the monitoring polygon is displaced from the calibration polygon by a smaller pixel displacement than the displacement threshold.
12 . The system of claim 1 , wherein the processor-executable instructions further cause the at least one processor to determine a calibration confidence score based on an image position of the calibration polygon to an image position of an optimal polygon, the optimal polygon representing an optimal position for the calibration polygon.
13 . The system of claim 12 , wherein the deviation threshold is variable based on the calibration confidence score.
14 . The system of claim 1 , wherein the first telematics data comprises inertial data indicative of vertical movement of the vehicle for the first period of time, and the suitability criteria includes the vertical movement of the vehicle being linear.
15 . The system of claim 1 , wherein the first telematics data comprises elevation data indicative of change in elevation of the vehicle for the first period of time, and the suitability criteria includes an elevation trajectory of the vehicle being linear.
16 . The system of claim 1 , further comprising at least one telematics sensor which captures the first telematics data and the second telematics data.