Invented by Soryoung KIM, LG Electronics Inc
The LG Electronics Inc invention works as follows
The disclosed method for controlling an autonomous vehicle includes: generating driving data by merging meta-information, including location-based information and image information, receiving an object detection algorithms selected based upon the meta-information, setting a path based on that object detection algorithm while monitoring the main object which has appeared at a location indicated by the location information and, when dangerous object information from a server is received, resetting the path to avoid a dangerous object. The present invention can include an autonomous vehicle, user terminals, and servers. These may work in conjunction with a UAV, Augmented Reality device (AR), Virtual Reality device (VR), and devices related to 5G services.Background for Control device for autonomous vehicle and method of controlling it
Field of Invention
The present invention is a method of controlling an autonomous vehicle as well as a device for controlling it. It also relates to the reduction of algorithm resources and computation times when detecting an object.
Related Art
Vehicles can be classified as internal combustion engines, external combustion engines, gas turbine vehicles or electric vehicles.
The development of an autonomous car capable of driving itself without the need for a human driver is currently in full swing.
To replace the human perception capabilities, different sensors, such as a sensor infrared, radar, camera, etc., are used. They are used. The camera can replace the human eye and capture an image of driving conditions of a car. The autonomous vehicle also analyzes the image captured by the camera, and then performs autonomous driving on the basis of the image analyzed.
The present invention is a method of control and a device for detecting an object more quickly in an image taken by an autonomous vehicle.
The present invention is aimed at providing a control method, and a device that can reduce the algorithm resources required by an autonomous vehicle to detect an object in an image.
The present invention is a method of controlling and a device that allows an image to be quickly analyzed and an algorithm to be reduced. It also makes it possible to handle an emergency efficiently.
The object detection algorithm is used to set a driving route while monitoring the main object in the place indicated by location information.
Meta information can be used to generate driving information that includes time and weather data.
The object detection algorithm can be used to detect an object in real-time that appears at a location indicated by the position information.
The object detection algorithm can be used to detect an object in real-time that appears at a location indicated by the information about the location and within a period of time indicated by information about the time.
The object detection algorithm can also include an algorithm to detect an auxiliary object that guides traffic rules observation while the vehicle is traveling.
The receiving of an object-detection algorithm can include the following: extracting the location data from driving information by a server; preparing a base station by the server in which each piece of location info is stored; searching the database by the client for an algorithm that matches the location data; and receiving the algorithm by the vehicle from the server.
The server can prepare the database in several steps: first, selecting an object-detection algorithm to detect the main object that was extracted in the first stage; second, matching the algorithm selected in the second phase with each piece received of location data from the other vehicle; and finally, storing the algorithm matched.
The server can also update the object detection algorithm using driving data from other vehicles.
The updating of the object-detection algorithm can include: detecting and tagging new objects based on driving information from other vehicles, updating the main object based upon the correlation between the driving path and the location information and the appearance of the objects using a deep learning method derived from the tagging data.
The server can search for object detection algorithms that are generated using the driving data received from other vehicles in real-time.
Click here to view the patent on Google Patents.