Mechanical arm grabbing method based on reference segmentation and spatial perception

CN120901933AActive Publication Date: 2025-11-07ZHEJIANG UNIV OF TECH
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202510870878.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-26
Publication Date
2025-11-07
Estimated Expiration
2045-06-26

AI Technical Summary

Technical Problem

Existing robotic arm grasping methods suffer from high computational costs, low grasping efficiency, and poor accuracy in complex environments. Existing 3D-RES methods also lack sufficient recognition accuracy under complex semantic and spatial ambiguity conditions, limiting their application in grasping tasks.

Method used

A robotic arm grasping method based on reference segmentation and spatial awareness is adopted. By acquiring point cloud data and natural language commands, the grasping posture result set is generated using the reference segmentation and spatial awareness model. This includes preprocessing of point cloud data, preprocessing of natural language commands, operation of the reference segmentation and spatial awareness model, generation of semantic mask and grasping candidate point set, and selection of optimal grasping posture.

Benefits of technology

While maintaining low computational resource consumption, it improves the robotic arm's ability to understand semantic targets and its grasping success rate in complex environments, achieving accurate, efficient and robust grasping posture prediction performance, and is suitable for semantic-driven human-machine collaboration, service robots and multi-object operation scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120901933A_ABST
    Figure CN120901933A_ABST
Patent Text Reader

Abstract

The invention discloses a mechanical arm grabbing method based on reference segmentation and spatial perception, which comprises the following steps: S1, acquiring point cloud data and a natural language instruction in a mechanical arm grabbing scene, and preprocessing to extract corresponding modal features; s2, inputting the modal features of the point cloud data and the modal features of the natural language instruction into a reference segmentation and spatial perception model to obtain a capture attitude result set; and S3, one optimal grabbing posture in the grabbing posture result set is selected as a control instruction, and the mechanical arm is driven through the control instruction to drive the clamping jaw to grab the target object. According to the method, the understanding ability and the grabbing success rate of the mechanical arm on a semantic target in a complex environment can be effectively improved while low computing resource consumption is kept, and accurate, efficient and robust grabbing posture prediction performance is achieved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Mechanical arm grabbing method driven by natural language

    CN117773920A

  • Object attitude estimation method and device, mechanical arm grabbing method, system and equipment

    CN118628563A

  • Grabbing and sampling method based on simple semantic point cloud

    CN119188728A

  • Robot grabbing detection method based on visual language action multi-mode alignment strategy

    CN119526405A

  • Object segmentation based on detected object-specific visual cues

    US9327406B1