Patents by Inventor Laiwei JIANG

Laiwei JIANG has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).

  • Multi-target tracking method based on feature decoupling with joint detection and feature re-identification
    Patent number: 12579663
    Abstract: A multi-target tracking method with joint detection and re-identification based on feature decoupling in the field of multi-target tracking for solving low tracking accuracy problem caused by conflicts in optimization directions in the joint training process of detection and re-identification tasks and improving tracking performance, includes the steps of: preprocessing multi-target tracking dataset, extracting feature F from the processed images and labels by the backbone network; decoupling feature F in the channel and spatial dimensions to obtain feature Fdet suitable for detection tasks and feature Fid suitable for re-identification tasks respectively, obtaining the position of the target detection bounding box in the image based on Fdet subdivision detection branch; calculating the similarity matrix D based on Fid and IOU overlap level, and processing matching of the target detection bounding box in the image with the trajectory already existed by using D and Hungarian algorithm to finally obtain the tra
    Type: Grant
    Filed: December 18, 2024
    Date of Patent: March 17, 2026
    Inventors: Laiwei Jiang, Ce Wang
  • Multi-Target Tracking Method Based on Feature Decoupling with Joint Detection and Feature Re-identification
    Publication number: 20260044966
    Abstract: A multi-target tracking method with joint detection and re-identification based on feature decoupling in the field of multi-target tracking for solving low tracking accuracy problem caused by conflicts in optimization directions in the joint training process of detection and re-identification tasks and improving tracking performance, includes the steps of: preprocessing multi-target tracking dataset, extracting feature F from the processed images and labels by the backbone network; decoupling feature F in the channel and spatial dimensions to obtain feature Fdet suitable for detection tasks and feature Fid suitable for re-identification tasks respectively, obtaining the position of the target detection bounding box in the image based on Fdet subdivision detection branch; calculating the similarity matrix D based on Fid and IOU overlap level, and processing matching of the target detection bounding box in the image with the trajectory already existed by using D and Hungarian algorithm to finally obtain the tra
    Type: Application
    Filed: December 18, 2024
    Publication date: February 12, 2026
    Inventors: Laiwei JIANG, Ce WANG