首页|基于YOLO-RepGFPN模型的煤矸检测方法研究

基于YOLO-RepGFPN模型的煤矸检测方法研究

扫码查看
现有基于深度学习的煤矸识别方法在光线昏暗、高噪声、遮挡等因素的干扰下容易出现漏检、错检的现象.针对该问题,提出一种基于YOLO-RepGFPN模型的煤矸检测方法.YOLO-RepGFPN模型基于YOLOV5s模型改进:1)在Backbone部分中引入CA注意力机制,使网络能充分利用丰富的上下文信息,增强特征学习能力,2)在Neck部分采用RepGFPN网络模块,增强模型对煤矸高级语义和低级空间特征的提取和融合能力,从而提高检测精度,3)修改损失函数为XIOU增强目标匹配和目标边框回归过程中的定位精度和鲁棒性.实验结果表明:RepGFPN煤矸识别方法与 目前主流YOLO算法相比识别精度(mAP)最高,达到94.7%,可有效避免恶劣条件下,煤矸识别时容易出现的漏检、误检和重检现象.
Research on Coal Gangue Detection Method Based on YOLO-RepGFPN Model
The existing coal gangue identification methods based on deep learning are prone to miss and misdetect under the interference of dim light,high noise,occlusion and other factors.In order to solve this problem,a coal gangue detection method based on YOLO-RepGFPN model was pro-posed.Based on the YOLOV5s model,the YOLO-RepGFPN model is improved:1)The CA attention mechanism is introduced into the Backbone part to make the network make full use of the rich context information and enhance the feature learning ability.2)the RepGFPN network module is used in the Neck part to enhance the model's ability to extract and fuse the high-level semantics and low-level spatial features of gangue,so as to improve the detection accuracy.3)Modify the loss function to XIOU to enhance the positioning accuracy and robustness in the process of target matching and target border regression.The experimental results show that the RepGFPN gangue identification method has the highest recognition accuracy(mAP)of 94.7%compared with the current mainstream YOLO algo-rithm,which can effectively avoid the phenomenon of missed detection,false detection and re-detec-tion that are easy to occur in the identification of coal gangue under harsh conditions.

coal gangue identificationdeep learningattention mechanismsRepGFPNloss function

陈森森、程刚、王龙腾

展开 >

安徽理工大学机械工程学院,安徽淮南 232001

煤矸识别 深度学习 注意力机制 RepGFPN 损失函数

安徽省高校协同创新项目

ZY7092021004

2024

佳木斯大学学报(自然科学版)
佳木斯大学

佳木斯大学学报(自然科学版)

影响因子:0.159
ISSN:1008-1402
年,卷(期):2024.42(3)
  • 9