METHODS AND SYSTEMS FOR GENERATING TRAINING DATA FOR DETECTING HORIZON AND ROAD PLANE Russian patent published in 2022 - IPC B60K35/00 G06K9/00 G06T17/00 G05D1/02 B60W40/08 B60W60/00 

FIELD: computer technology; machine learning.

SUBSTANCE: invention relates to computer methods and systems for machine learning. A method for forming a machine learning model for use in an unmanned vehicle in order to predict the road plane based on an image includes generation of a training data set in a computer by receiving data collected during the operation of a training vehicle on the road, including an image related to the first location of the training vehicle on the road, and odometrical data related to the second location of the training vehicle on the road. Generation of the training data set also includes determination of the road plane in the second location of the training vehicle based on odometrical data, determination of a line on the image, addition of this line to the image to form a training example, and addition of this training example to the training data set. This training data set is provided to a machine learning algorithm performed in the computer to form a machine learning model used in the unmanned vehicle, and a new road plane is determined based on a new image using the machine learning model.

EFFECT: correct determination of the road plane is achieved.

20 cl, 9 dwg