Three-dimensional laser point cloud data preprocessing method and device

Abstract

The invention provides a three-dimensional laser point cloud data preprocessing method and apparatus. The method comprises the steps of obtaining three-dimensional laser point cloud collected by a collection vehicle, track information when the collection vehicle collects the three-dimensional laser point cloud, reflection intensity information and height information of the three-dimensional laserpoint cloud, and height information in the track information of the collection vehicle; projecting the three-dimensional laser point cloud and the track information of the acquisition vehicle onto a preset plane; filtering the three-dimensional point cloud by using height information in the track information of the acquisition vehicle and a preset concerned height range; and generating an intensity graph and a height graph covering the whole scene according to the reflection intensity information and the height information of the filtered three-dimensional point cloud and the projection graphof each point on the preset plane. According to the method, the efficiency of processing the laser point cloud data can be improved, and images of objects which are not concerned in the environment corresponding to the three-dimensional laser point cloud can be filtered.

Description

technical field [0001] The invention relates to the technical field of information processing, in particular to a method and device for preprocessing three-dimensional laser point cloud data. Background technique [0002] In the production process of high-precision maps, the existing mainstream technical process mainly includes five basic links: data collection, data preprocessing, automatic identification, manual verification and correction, compilation and release. [0003] Among them, the preprocessing stage mainly completes the time-space correlation and data fusion of the collected discrete pictures and 3D laser point cloud data, and then generates complete data covering the entire road scene, serving the subsequent automatic identification and manual inspection links. [0004] At present, the data output in the preprocessing stage mainly include two mainstream data types: continuous video images with spatio-temporal information and 3D laser points of scenes after splic...

AI Technical Summary

Problems solved by technology

[0005] Although the current machine learning algorithm represented by deep learning has achieved fruitful results in the field of image recognition, since the recognition algorithm is applied to a single two-dimensional image, only after complex three-dimensional mapping and fusion processes can these two-dimensional The recognition result is converted into a 3D scene map, which increases the complexity of the algorithm on the one hand, and on the other hand, it is easy to introduce additional errors in the conversion process, which affects the final mapping effect

For the 3D laser points of the scene output by the existing technology, the current point cloud recognition algorithm based on deep learning still has a certain gap compared with the image recognition algorithm in terms of recognition efficiency and recognition accuracy.

In addition, in the process of manual inspection, the 3D interactive environment based on point cloud data is more complicated than the 2D interactive environment based on images, which is not convenient for the improvement of operation efficiency

Images