Researching | Unlabeled Video Retrieval Method of Mining Personnel Based on MK-YOLOV4

Laser & Optoelectronics Progress, Volume. 59, Issue 4, 0415003(2022)

Unlabeled Video Retrieval Method of Mining Personnel Based on MK-YOLOV4

Yunhui Zhao¹, Xiaozhou Cheng², Kaiwen Dong^1、*, Xiao Yun¹, Yanjing Sun¹, and Yingjie Han¹

Author Affiliations

¹School of Information and Control Engineering, China University of Mining and Technology, Xuzhou , Jiangsu 221116, China

²Sinosteel Maanshan Institute of Mining Research Co., Ltd., Maanshan, Anhui 243000, China

show less

Abstract Get PDF(in Chinese)

Figures&Tables (11)

References (23)

Paper Information

Figures & Tables(11)

Fig. 1. MK-YOLOV4 network structure

Download full size

Fig. 2. Flowchart of anchor box generation method

Download full size

Fig. 3. CA-NET structure

Download full size

Fig. 4. Data enhancement effect. (a) Color jitter; (b) random erasure

Download full size

Fig. 5. Some examples of Miner-Market personnel re-identification data set

Download full size

Fig. 6. Algorithm performance comparison

Download full size

Fig. 7. Detection results in actual mining scene

Download full size

Fig. 8. Weight coefficient $β$ analysis. (a) mAP; (b) Rank1

Download full size

Fig. 9. Mining staff re-identification search results

Download full size

Table 1. Comparison of different network performance
View table
Table 1. Comparison of different network performance
Backbone mAP /% Rank1 /%
ResNet50^［23］ 65.6 82.7
SeResNet50^［20］ 68.4 84.5
ResNet50-IBN-a^［19］ 67.0 83.9
CA-NET 69.0 85.6

Table 2. Performance analysis of different models in Miner-Market data set
View table
Table 2. Performance analysis of different models in Miner-Market data set
Model mAP /% Rank1 /%
CA-NET+Trihardwc loss+Color jitter 78.6 91.2
CA-NET+Trihardwc loss+Color jitter+REA 82.4 93.0

Tools

Get Citation

Copy Citation Text

Yunhui Zhao, Xiaozhou Cheng, Kaiwen Dong, Xiao Yun, Yanjing Sun, Yingjie Han. Unlabeled Video Retrieval Method of Mining Personnel Based on MK-YOLOV4[J]. Laser & Optoelectronics Progress, 2022, 59(4): 0415003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Mar. 1, 2021

Accepted: Apr. 2, 2021

Published Online: Feb. 15, 2022

The Author Email: Kaiwen Dong (dongkaiwen1996@outlook.com)

DOI:10.3788/LOP202259.0415003

Topics

laser devices and laser physics

Lasers and Laser Optics

laser manufacturing

Instrumentation, Measurement and Metrology