| CPC G05D 1/0088 (2013.01) [G05D 1/0287 (2013.01); G06N 20/00 (2019.01); G06T 7/292 (2017.01); G06T 7/593 (2017.01); G06T 7/80 (2017.01); G06T 7/85 (2017.01); H04N 13/246 (2018.05); G05D 2201/0213 (2013.01); G06T 2207/10012 (2013.01)] | 18 Claims |
|
1. A method comprising:
receiving sensor data representing an image associated with a driving environment of a vehicle;
determining an object within the driving environment, wherein determining the object comprises automatically drawing a box over the object in the image, a side of the box is placed on each of a top, a bottom, a first side and a second side boundary of the object and the box delineates the extents of the object within the image;
determining if the object is a second vehicle:
if the object is a second vehicle, changing an autonomous driving mode to follow the second vehicle automatically, wherein changing the autonomous driving mode comprises:
determining, for each image in a series of images, a two-dimensional center of the box based on a crossing point of a first diagonal, from a first corner of the box to a third corner of the box, and a second diagonal, from a second corner of the box to a fourth corner of the box;
determining if the second vehicle is turning;
if the vehicle is turning, adjusting the center of the box to compensate for the second vehicle turning, which maintains the vehicle within a lane of a road;
determining a two-dimensional vector associated with a change in a location of the center of the box between subsequent images;
maintaining a position relative to the second vehicle by changing a steering angle and/or a velocity to reverse the two-dimensional vector; and
if the object is not a second vehicle, changing the autonomous driving mode to follow the lane viewed in the driving environment.
|