公开(公告)号：US20250061701A1

公开(公告)日：2025-02-20

申请号：US18806740

申请日：2024-08-16

Applicant: China University of Mining and Technology

Inventor： Xiaohu ZHAO ,  He TIAN ,  Peng YOU ,  Tingyu CHE

IPC: G06V10/82 ,  G01N25/72 ,  G01N33/24 ,  G06T7/00 ,  G06V10/26 ,  G06V10/30 ,  G06V10/52 ,  G06V10/77 ,  G06V10/80

Abstract: A detection method for infrared thermal image damage area of coal rock includes: step 1: collecting and recording the infrared thermal image of coal rock in the process of loading destruction; step 2: processing a gray-scale transformation on the collected infrared thermal image of coal rock; step 3: denoising an infrared thermal image of coal rock after the gray-scale transformation by the dense residual image denoising algorithm of autocorrelation network; step 4: conducting an area segmentation on the infrared thermal image of coal rock after denoising to extract eigenvalues by using a coal-rock infrared thermal image damage area segmentation algorithm of improved encoder-decoder network; step 5: observing the damage area of coal rock. The detection method improves the denoising effect of the infrared thermal image of coal rock, strengthens the extraction of the characteristics of the damage area, and improves the accuracy of the segmentation.

公开(公告)号：US20250022165A1

公开(公告)日：2025-01-16

申请号：US18755735

申请日：2024-06-27

Applicant: CHINA UNIVERSITY OF MINING AND TECHNOLOGY

Inventor： Xiaohu ZHAO ,  Xingyi YOU ,  Peng YOU ,  Sheng YE ,  Yong LIU

IPC: G06T7/73 ,  G06T5/10 ,  G06T5/20 ,  G06V10/44 ,  G06V10/80

Abstract: The present invention discloses an interactive behavior understanding method for posture reconstruction based on skeleton and image features. The steps are as follows: constructing and preprocessing the data set, extracting skeleton and image features, fusing and reconstructing these features, and conducting experimental evaluation and validation. This method retains the purity of skeleton features for human behavior information extraction and uses image features to retain effective environmental information, complementing the model feature information. Skeleton features are extracted using a graph convolution network, enhancing the relevance of input skeleton point information for accurate feature extraction. Effective image features are quickly and accurately extracted through the Vision Transformer network combined with a multi-head attention mechanism.