Patent — Method and system for prediction of objects in the space in real-time via 3D replica

Filed Jun 7, 2021

The patent relates to a specific way of understanding objects and predicting their movement by analyzing data obtained from camera images. Specifically, the objects in question are those that can move, any means of transportation, their environment, such as buildings, roads, parking spaces, etc.

Currently available solutions for understanding the objects in images captured by a camera rely on data directly obtained from the neural network. This approach results in incomplete and imprecise data that have limited usefulness.

This problem is solved and the understanding of the space and objects in the image captured by the camera is significantly improved, as well as the prediction of the movement of the objects being viewed, in a way that provides innovation by using data from the image to make a vector model of the object using the neural network, and then placing the obtained data in the database, after which it is used to create a virtual simulation that represents a controlled digital replica of the viewed space in 3-dimensions (3D) through intelligent control that includes physics models for additional realism and accuracy checks of the acquired data.

The accuracy is improved with additional intelligence in the prediction process so that real-life situations where the objects being viewed are located in space, how they move, and whether it is possible for the situation to change, are taken into account through the physics model.