Processing a point cloud generated by an environment detection device of a vehicle into input data for a neural network

摘要

The invention relates to a method for processing a point cloud (8) generated by an environment detection device (2) of a vehicle (1) and representing an environment (6) of the vehicle (1) into input data (12) for a neural network (14). , comprising the steps a) receiving a point cloud (8) from an environment detection device (2) of a vehicle (1), b) specifying a determination distance, c) selecting an unprocessed point (10) of the provided point cloud (8), d) determining, whether the selected point (10) is a core point for which there is a core distance within which it has a predetermined minimum number of adjacent points (10), the core distance being less than or equal to the determination distance if the selected one Point (10) is a key point, determining a respective reachability distance to neighboring points (10) located within the determination distance, f) forming an order of processed points (10) by taking up the neighboring point (10) with the shortest reach distance as the next point (10) in the order and setting this point (10) as the next selected point, g) repeating steps d) to f), to all points (10) of the point cloud (8) are processed, the order of the processed points (10), which results after processing all points (10) of the point cloud (8), together with the accessibility distances associated with the points (10) Forms input data (12) for a neural network (14). The invention further relates to a computing device which is designed to carry out such a method and has a neural network, and to a vehicle having such a computing device.