A tower crane operating environment perception method based on multi-scale multi-modal feature fusion decoding

By employing a multi-scale, multi-modal feature fusion decoding method, utilizing 4D sparse point cloud and video image data, and combining a Transformer architecture and a multi-scale decoder, the high cost and insufficient accuracy of tower crane operating environment perception are addressed, achieving efficient and accurate environmental perception capabilities.

CN121353985BActive Publication Date: 2026-06-26SHANXI JIANTOU YUNSHUZHI TECH CO LTD
2 Cites 0 Cited by

Patent Information

Application Number
CN202511563427.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-10-30
Publication Date
2026-06-26
Estimated Expiration
2045-10-30

AI Technical Summary

Technical Problem

Existing tower crane operating environment perception technologies suffer from high costs, poor applicability, and insufficient accuracy in complex construction scenarios. In particular, image analysis methods based on machine vision struggle to acquire three-dimensional spatial information, while methods based on lidar are greatly affected by weather conditions and are also costly.

Method used

A multi-scale, multi-modal feature fusion and decoding method is adopted, which uses 4D sparse point cloud data and video surveillance image data to collect sparse point cloud data and video images to collect video images. Combined with the Transformer multi-modal fusion architecture with deformable multi-head attention mechanism and multi-scale decoder, the efficient fusion and decoding of data is achieved.

Benefits of technology

While reducing costs, it significantly improves the efficiency and accuracy of tower crane operating environment perception, adapts to severe weather conditions, and solves the limitations and insufficient accuracy of existing technologies.

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

Abstract

The present application relates to a tower crane operating environment perception method based on multi-scale multi-modal feature fusion decoding, belonging to the technical field of intelligent construction, specifically comprising that the method is mainly based on sparse point cloud data collected by 4D imaging radar and video image data collected by monitoring camera, through twice feature extraction of fusion of two modal data based on deformable attention mechanism Transformer architecture, while the two modal data are down-sampled, the above feature extraction process is repeated to realize the fusion feature information of multi-modal data from different scales, and finally the multi-scale decoding of the fusion feature information of different scales is carried out, so that the final tower crane environment perception result is obtained, the present application carries out multi-modal fusion analysis of 4D sparse point cloud data and video monitoring image data in different scales, and then the recognition efficiency and accuracy of tower crane operating environment perception problem are significantly improved.
Need to check novelty before this filing date? Find Prior Art