A virtual lidar data generation method based on a virtual world

Abstract

The invention relates to a virtual lidar data generation method based on a virtual world, comprising the following steps: step S100, acquiring scene information of the virtual world; step S200, analyzing the scene cache information obtained in step S100 to generate scene point cloud data; step S300, analyzing the scene target information obtained in step S100, and generating a scene target point cloud index set; S400, downsampling the scene point cloud data to compose the true value data to generate the virtual lidar data. The invention provides a lidar point cloud data source containing truevalue data for the development of an autopilot sensing algorithm based on a lidar sensor, point cloud data can be collected efficiently without any restrictions and the algorithm development processcan be sped up; at the same time, the consumption of human time costs caused by data acquisition and annotation in the development process is greatly reduced.

Description

technical field [0001] The invention relates to the application field of unmanned aerial vehicle simulation computing, in particular to a virtual laser radar data generation method based on a virtual world. Background technique [0002] Unmanned driving simulation technology, as an indispensable link in the development of unmanned driving, is mainly used for the development of various algorithms for unmanned driving, dynamic model verification, functional safety verification, etc. The driverless algorithm integrates environmental perception, road modeling, path planning, decision control, and functional safety design, involving various sensors, body control units, and body ECUs. For the environment perception algorithm, it involves monocular camera, binocular camera, ultrasonic radar, millimeter-wave radar, lidar, etc. for perceiving the environment. Based on the raw data of the sensor, the driving environment of the vehicle is modeled to realize the perception of the enviro...

AI Technical Summary

Problems solved by technology

[0003] The current environmental perception algorithm is still in the research and development stage, and the development of the algorithm is heavily dependent on the raw data of the sensor. At present, the raw data of the sensor is mainly collected through various sensors equipped with real vehicles on the road. This method is costly and has great limitations. The collection area is limited and the collection takes a long time

Since the existing technology cannot simulate the original lidar data in the simulation environment, it is impossible to use the simulation environment to accelerate the development of lidar point cloud data algorithms

Images