SLAE-Projected GANs的煤矿井下图像生成方法  

Image generation method of coal mine underground based on SLAE-Projected GANs

作　　者：高志军 孙丽丽 彭重霄 孙银焕 Gao Zhijun;Sun Lili;Peng Chongxiao;Sun Yinhuan(School of Computer&Information Engineering,Heilongjiang University of Science&Technology,Harbin 150022,China)

机构地区：[1]黑龙江科技大学计算机与信息工程学院,哈尔滨150022

出　　处：《黑龙江科技大学学报》2025年第1期153-159,共7页Journal of Heilongjiang University of Science And Technology

基　　金：黑龙江省省属高等学校基本科研业务费项目(2024-KYYWF-1082,2022-KYYWF-0565)。

摘　　要：针对煤矿井下复杂多变的环境而无法获取高质量的煤矿井下图像的问题,提出了一种基于SLAE-Projected GANs的煤矿井下图像生成方法。采用Projected GANs网络框架,向该网络的上采样结构中注入噪声,保证模型生成图像过程中的稳定性,引入空间注意力机制,增强模型的表征能力从而有效地提取图像的特征信息。结果表明,文中的方法生成图像的峰值信噪比、结构相似性和Frechet Inception距离分别为19.065、0.629和30.022,优于其他现有图像生成算法,能够生成更加稳定、高质量和高分辨率的煤矿井下图像,有助于图像的目标检测和图像识别。This paper seeks to address the problem of being unable to obtain high-quality coal mine underground images due to the complex and changing environment,and proposes a coal mine underground image generation method based on SLAE-Projected GANs.The study works by adopting the framework of Projected GANs network and injecting noise into the up-sampling structure of the network to ensure the stability of the model in the process of generating images;and introducing the spatial attention module to enhance the model′s characterization ability so as to effectively extract the feature information of the image.The results show that PSNR,SSIM and FID of the image generated by this method are 19.065,0.629 and 30.022,respectively,which are better than other existing image generation algorithms,and can generate more stable,high-quality and high-resolution images of coal mine underground,which can help the target detection and image recognition of images.

关 键 词：煤矿井下图像 图像生成 空间注意力机制 SLAE-Projected GANs 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]