A three-dimensional scene reconstruction method, device and electronic device based on lidar

Abstract

The invention provides a three-dimensional scene reconstruction method, device and electronic equipment based on laser radar, and relates to the technical field of computer vision. Based on joint detection results, an implicit expression function of an object is learned, and a compact object surface is reconstructed, so as to realize accurate and accurate reconstruction of the three-dimensional scene. complete understanding. The method includes: acquiring point cloud data of a target area; jointly detecting the point cloud data by a neural network based on an attention mechanism to obtain a joint detection result; the joint detection result includes objects in the target area and the layout of the target area; learning based on the joint detection result Obtain the implicit expression function of the object in the target area; reconstruct the 3D scene of the object in the target area based on the implicit expression function. The three-dimensional scene reconstruction device based on lidar is applied to a three-dimensional scene reconstruction method based on lidar. The three-dimensional scene reconstruction method based on lidar is applied in electronic equipment.

Description

technical field [0001] The present invention relates to the technical field of computer vision, and more particularly, to a method, device and electronic device for three-dimensional scene reconstruction based on laser radar. Background technique [0002] Three-dimensional scenes include indoor scenes and outdoor scenes. In the prior art, the application number is CN201910925338.0, an octree-based robot vision-guided three-dimensional object reconstruction method, using an octree structure as a storage structure for point cloud data, and selecting As a reconstruction algorithm, the fixed region growing method discloses the detection and reconstruction of objects alone. The application number is CN201811466786.0, a three-dimensional reconstruction method of indoor scenes based on RGB-D images, which uses semantic segmentation results to repair depth image holes, provides object outline and category information for three-dimensional reconstruction, and obtains the shape and sh...

AI Technical Summary

Problems solved by technology

[0003] In the existing technology, most rely on 2D images and multi-frame data for reconstruction, and there are problems such as high hardware requirements, complex algorithms, and long processing time

More importantly, the reconstruction of the existing technology is mostly limited to the detection of objects alone or the prediction of the 3D scene layout alone, and there are few collaborative detections of the two, which have certain limitations in time efficiency and network models.

Images