A space positioning method, device, apparatus and storage medium

By constructing a neural radiation field and combining 3D point cloud data with user behavior data, the problems of insufficient positioning accuracy and high computational resource consumption in existing technologies are solved, realizing an efficient and robust spatial positioning method and improving the user experience.

CN121033158BActive Publication Date: 2026-09-11LINGBAN INTELLIGENT (HANGZHOU) INFORMATION TECHNOLOGY CO LTD
4 Cites 0 Cited by

Patent Information

Application Number
CN202511100733.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-08-07
Publication Date
2026-09-11
Estimated Expiration
2045-08-07

AI Technical Summary

Technical Problem

Existing technologies for 3D spatial mapping and positioning suffer from insufficient accuracy, susceptibility to changes in lighting and dynamic object occlusion, high computational resource consumption, and complex operation procedures, making it difficult to meet the needs of high-precision scenarios and reduce implementation costs.

Method used

By acquiring 3D point cloud data and multi-view discrete images of the target scene, a neural radiation field is constructed. Effective poses are extracted by combining user behavior data, a set of rendered images is generated, and pose correction is performed based on the 3D point cloud data to achieve fast matching and efficient positioning.

Benefits of technology

It improves positioning accuracy, reduces computational cost, enhances positioning robustness and user experience, and meets the requirements for high-precision positioning.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The application provides a space positioning method, device and equipment and a storage medium. The method comprises the following steps: constructing a neural radiation field based on three-dimensional point cloud data and multi-view discrete images of a target scene; obtaining user behavior data, including a basic pose image shot by a user and / or editing data when a virtual object is edited, extracting effective poses therefrom and generating a virtual shooting pose set; generating a rendering image set of the bound poses through the neural radiation field; and finally receiving a positioning request, combining a to-be-positioned image, the rendering image set and the three-dimensional point cloud data to determine a target pose. The application determines the target pose in combination with multi-modal data, breaks through the precision limitation of a single-modal sensor in a complex scene, eliminates visual matching cumulative errors and dynamic environment interference, improves positioning precision while reducing the consumption of computing resources, extracts effective poses in combination with user behavior data for rendering, and avoids the problems of complex calculation or inaccuracy caused by redundant rendering.
Need to check novelty before this filing date? Find Prior Art