Systems and methods for trailer hitch ball position and hitch angle

Abstract

The disclosure is generally directed to systems and methods for trailer hitch ball position location including receiving a plurality of image frames from a camera directed at a front of a trailer coupled to a vehicle at a coupling point, modeling the image frames in a convolutional neural network to form an initial estimate of a pivot point position as a location, optimizing the model using a nonlinear equation to identify the pivot point position, and locating the coupling point as the optimized pivot point position. The convolutional neural network includes a plurality of bounding boxes centered as locations of predetermined markers on the front of the trailer, the modeling via a reverse pinhole projection of a pixel onto a three-dimensional coordinate frame projected on a trailer plane to enable computed marker positions to determine relative dimensions of the trailer and locate the coupling point. The optimizing includes solving a nonlinear least-squares (NLLSQ) optimization formulation for multiple markers on the trailer. The method include the use of geometrical means to determine the angle of articulation between the trailer and tow-vehicle (hitch angle) by tracking the plurality of bounding boxes centered as locations of predetermined markers on the front of the trailer, as the vehicle and trailer move in a circular arc with respect to each other.