Method, device, apparatus and medium for extracting obstacle perception error data

Abstract

The embodiment of the invention discloses a method, a device, an apparatus and a medium for extracting obstacle perception error data. The method comprises the following steps of: acquiring a road test data set of a vehicle, wherein the road test data set comprises sensor data for sensing obstacles, obstacle perception result data and artificial driving behavior data, obtaining example data of obstacle false identification and / or obstacle leakage identification by comparing and analyzing the obstacle perception result data and the artificial driving behavior data; wherein the example data includes sensor data. The technical scheme extracts a large amount of corresponding instance data such as obstacle leakage identification and / or error identification in the driving process of a vehicle during road measurement, provides data support for the retraining of obstacle perception model, which can optimize the obstacle perception model, improve the accuracy and reliability of obstacle recognition, and reduce the potential safety hazards of unmanned vehicles.

Description

technical field [0001] The embodiments of the present invention relate to the field of unmanned driving technology, and in particular to a method, device, device and medium for extracting obstacle perception error data. Background technique [0002] In the unmanned driving perception system, the perception and recognition of obstacles mainly rely on the image data output by the camera, the point cloud data output by the lidar, and the data output by the radio radar. In the prior art, in the iterative development stage of the algorithm, a large amount of data collection and standards are usually carried out through the open-loop mode of manual driving, which is used for obstacle perception model training, so as to identify and classify obstacles through the obstacle perception model. [0003] However, due to the wide variety of obstacles, non-obstacles such as green plants, branches, and water on the ground are misidentified, and / or low obstacles such as traffic cones, tripod...

AI Technical Summary

Problems solved by technology

[0003] However, due to the wide variety of obstacles, non-obstacles such as green plants, branches, and water on the ground are misidentified, and / or low obstacles such as traffic cones, tripods, or children and pedestrians are sometimes missed. Occurs, bringing hidden dangers to the driving safety of unmanned vehicles

Images