Methods for determining and reporting vehicle following distance
Systems, methods, models, and training data for models are discussed, for determining vehicle positioning, and in particular identifying tailgating. Simulated training images showing vehicles following other vehicles, under various conditions, are generated using a virtual environment. Models are trained to determine following distance between two vehicles. Trained models are used in detection of tailgating, based on determined distance between two vehicles. Results of tailgating are output to warn a driver, or to provide a report on driver behavior. Following distance over time is determined, and simplified following distance data is generated for use at a management device.
1 . A method for generating following distance data over time for a subject vehicle, the method comprising:
accessing a set of images, each image of the set of images being from a perspective of the subject vehicle and including a respective representation of another vehicle positioned in front of the subject vehicle, and each image of the set of images being associated with a respective time;
generating following distance data based on at least a subset of images in the set of images, comprising:
for each image in the subset of images, determining a respective physical distance between the subject vehicle and the other vehicle; and
compiling the following distance data over time as a plurality of following distance data points, each following distance data point representing the determined physical distance for a respective image in the subset of images and the time associated with the respective image in the subset of images;
generating simplified following distance data, comprising:
identifying select data points from the following distance data for inclusion in the simplified following distance data, based on differences between the data points and iteratively-defined reference lines through portions of the following distance data; and
compiling the select data points as the simplified following distance data, excluding data points which are not identified as select data points; and
outputting the simplified following distance data.
2 . The method of claim 1 , wherein identifying select data points from the following distance data for inclusion in the simplified following distance data, based on differences between the data points and iteratively-defined reference lines through portions of the following distance data, comprises, for each data point of the following distance data:
determining a minimum difference between the data point and a corresponding reference line of the iteratively-defined reference lines; and
identifying the data point as a select data point if the minimum difference between the data point and the corresponding reference line exceeds a difference threshold.
3 . The method of claim 1 , wherein identifying select data points from the following distance data for inclusion in the simplified following distance data, based on differences between the data points and iteratively-defined reference lines through portions of the following distance data, comprises:
defining a reference line through the following distance data;
identifying a candidate data point of the following distance data, where a minimum difference between the candidate data point and the reference line is greater than a minimum difference between other data points and the reference line; and
if the minimum difference between the candidate data point and the reference line exceeds a difference threshold:
identifying the candidate data point as a select data point for inclusion in the simplified following distance data; and
defining new reference lines which intersect the candidate data point through the following distance data.
4 . The method of claim 3 , wherein identifying select data points from the following distance data for inclusion in the simplified following distance data, based on differences between the data points and iteratively-defined reference lines through portions of the following distance data, further comprises:
defining an additional reference line through the following distance data which intersects a select data point previously identified for inclusion in the simplified following distance data;
identifying an additional candidate data point of the following distance data, where a minimum difference between the additional candidate data point and the additional reference line is greater than a minimum difference between other data points and the additional reference line; and
if the minimum difference between the additional candidate data point and the additional reference line exceeds the difference threshold: identifying the additional candidate data point as an additional select data point for inclusion in the simplified following distance data.
5 . The method of claim 1 , wherein:
the following distance data comprises at least one data series of following distance data points;
identifying select data points from the following distance data for inclusion in the simplified following distance data, based on differences between the data points and iteratively-defined reference lines through portions of the following distance data, comprises, for each data series in the following distance data, until all data points in each data series are within a distance threshold of a corresponding reference line:
identifying a first data point in the data series and a last data point in the data series as select data points for inclusion in the simplified following distance data;
defining a reference line between the first data point in the data series and the last data point in the data series;
identifying a candidate data point between the first data point in the data series and the last data point in the data series, where a minimum difference between the candidate data point and the reference line is greater than a minimum difference between other data points in the data series and the reference line;
if the minimum difference between the candidate data point and the reference line exceeds a difference threshold:
identifying the candidate data point as a select data point for inclusion in the simplified following distance data;
defining a component data series between the first data point in the data series and the candidate data point; and
defining another component data series between the candidate data point and the last data point in the data series; and
if the minimum difference between the candidate data point and the reference line exceeds a difference threshold:
excluding each data point of the data series between the first data point and the last data point from inclusion in the simplified following distance data.
6 . The method of claim 1 , further comprising:
accessing a plurality of images captured by an image capture device positioned at the subject vehicle; and
identifying the set of images as images of the plurality of images which include another vehicle positioned in front of the subject vehicle.
7 . The method of claim 1 , further comprising:
accessing a plurality of images captured by an image capture device positioned at the subject vehicle; and
identifying the set of images as images of the plurality of images which include another vehicle positioned in front of the subject vehicle and in a common lane of travel with the subject vehicle.
8 . The method of claim 1 , wherein, for each image in the subset of images, determining the respective physical distance between the subject vehicle and the other vehicle comprises: applying a following distance determination model to the respective image, the following distance determination model being a machine learning model trained based on minimization of a following distance loss function for a training set of images representing a respective lead vehicle from a perspective of a following vehicle associated with a label indicating following distance between the lead vehicle and the following vehicle.
9 . The method of claim 1 , wherein, for each image in the subset of images, determining the respective physical distance between the subject vehicle and the other vehicle comprises:
determining at least one image attribute based on a positional measure between the other vehicle as represented in the respective image and at least one boundary of the image; and
applying a following distance determination model to determine a following distance based on the determined at least one image attribute.
10 . The method of claim 1 , wherein, for each image in the subset of images, determining the respective physical distance between the subject vehicle and the other vehicle comprises:
determining, by at least one processor, a first vertical position in the image representing a bottom of the other vehicle;
accessing, by the at least one processor, a second vertical position in the image representing a static physical distance from the subject vehicle;
determining a transformed first vertical position by applying, by the at least one processor, an image transformation matrix to the first vertical position;
determining a transformed second vertical position by applying, by the at least one processor, the image transformation matrix to the second vertical position;
determining an image distance between the transformed first vertical position and the transformed second vertical position; and
determining the following distance as a physical distance between the other vehicle and the subject vehicle based on the determined image distance and the static physical distance.
11 . The method of claim 1 , wherein outputting the simplified following distance data comprises transmitting, by a communication interface positioned at the subject vehicle, the simplified following distance data to a device remote from the subject vehicle.
12 . The method of claim 11 , further comprising presenting, by a user interface of the device remote from the vehicle, at least a portion of the following distance data.
13 . The method of claim 1 , further comprising, for each data point in the simplified following distance data:
determining whether the respective physical distance between the subject vehicle and the other vehicle for the data point is within a tailgating distance criteria;
identifying that the subject vehicle is not tailgating the other vehicle at a time corresponding to the data point if the respective physical distance is outside of the tailgating distance criteria; and
identifying that the subject vehicle is tailgating the other vehicle at a time corresponding to the data point if tailgating criteria are met, wherein the tailgating criteria includes the respective physical distance being within the tailgating distance criteria.
14 . The method of claim 13 , further comprising, in response to an identification that the subject vehicle is tailgating the other vehicle, outputting a tailgating indication.
15 . The method of claim 14 , wherein outputting the tailgating indication comprises transmitting the tailgating indication to a management device.
16 . The method of claim 14 , further comprising:
receiving, at a management device, respective simplified following distance data for a plurality of subject vehicles;
receiving, at the management device, location data for the plurality of subject vehicles;
accessing, by the management device, respective tailgating indications for the plurality of subject vehicles; and
for each tailgating indication, associating a location of the respective subject vehicle as indicated in the location data, at a time of the tailgating indication, with a roadway segment corresponding to the location of the subject vehicle.
17 . The method of claim 16 , further comprising:
quantifying, by at least one processor of the management device, an amount of tailgating indications for at least one roadway segment of interest; and
identifying, by the at least one processor of the management device, at least one high-risk roadway segment, by identifying at least one roadway segment where an amount of associated tailgating indications exceeds a tailgating risk threshold.
18 . The method of claim 1 , further comprising:
receiving, by a management device, respective simplified following distance data for a plurality of subject vehicles;
receiving, by the management device, respective location data for the plurality of subject vehicles;
for each data point in the simplified following distance data for each subject vehicle of the plurality of subject vehicles: associating a location of the respective subject vehicle at a respective time of the data point with a roadway segment corresponding to the location of the respective vehicle at the respective time; and
quantifying following distance for the plurality of subject vehicles for at least one roadway segment of interest.