Semantic map incremental updating method and system

Abstract

The invention belongs to the technical field of map updating, and particularly relates to a semantic map incremental updating method and system. The method specifically comprises the following steps: constructing a current local semantic map; registering the current local semantic map and the previous global semantic map, and converting the current local semantic map and the previous global semantic map to the same coordinate system; detecting the semantic inconsistent region of the point cloud corresponding to the current local semantic map and the previous global semantic map to obtain a semantic change region; respectively calculating point cloud distribution categories of a previous global semantic map corresponding to the semantic change area and a current local semantic map, and comparing the point cloud distribution categories to determine an environment change area; and extracting the environment change area and updating the environment change area to a previous global semantic map to obtain a current global semantic map. According to the scheme, the change area is preliminarily detected through semantic comparison, and whether the change area is the environment change area or not is determined through comparison of the point cloud distribution categories, so that the accuracy of environment change detection is improved, and incremental updating of the semantic map is more complete.

Description

technical field [0001] The invention belongs to the technical field of map updating, and more particularly relates to a semantic map incremental updating method and system. Background technique [0002] The semantic map is a high-precision semantic map, which contains rich semantic information, which allows the automatic driving system to fully perceive the surrounding environment information, so as to make appropriate driving actions. [0003] Due to the large amount of semantic map data, the construction speed of the semantic map still cannot meet the real-time performance required by autonomous vehicles under the limitation of the performance of current hardware devices. The system then calls the semantic map in real time for vehicle control, so that the real-time control decision-making of the automatic driving system can be guaranteed. However, during the period from the completion of the semantic map to the call of the automatic driving system, the relevant environmen...

AI Technical Summary

Problems solved by technology

[0003] Due to the large amount of semantic map data, the construction speed of the semantic map still cannot meet the real-time performance required by autonomous vehicles under the limitation of the performance of current hardware devices. The system then invokes the semantic map in real time for vehicle control, so that the real-time control decision-making of the automatic driving system can be guaranteed

However, during the period from the completion of the semantic map to the call of the automatic driving system, the relevant environment may change. If the automatic driving directly uses the previous semantic map, it will seriously affect driving safety. The map is updated

[0004] Traditionally, there are two main ways to update the map. The first is full update, that is, the entire map is updated, but this is computationally intensive and the update speed is slow; the second is incremental update, that is, only the changed The area is updated, which generally determines the change area and updates it by matching the features of the point cloud data, so that the calculation amount is smaller than the full update and the update speed is fast, but for the huge amount of semantic information, the point cloud feature matching The amount of calculation is still too large, so it is not suitable for the incremental update of the semantic map

[0005] In the prior art, there is an incremental semantic map update method based on feature point detection and segmentation, which detects changes in the environment by comparing the semantic information of the point cloud and incrementally updates the semantic map. This method relies too much on semantic segmentation and can easily lead to failure The convenience of accurately changing the area makes the incremental update of the semantic map incomplete; moreover, the detection speed and accuracy of the map changing area in the existing technical solutions are relatively low. The effect of volume update is the current technical difficulty

Images