基于U-Net的康普顿相机成像算法研究  

Research on U-Net Based Imaging Algorithm for Compton Camera

作　　者：宋耀洲 徐伟 卢棚 金龙泉 赵丽文 SONG Yaozhou;XU Wei;LU Peng;JIN Longquan;ZHAO Liwen(Anhui University of Science and Technology,Huainan 232001,China;Institute of Energy,Hefei Comprehensive National science Center,Anhui,Hefei 230031,China)

机构地区：[1]安徽理工大学计算机科学与工程学院,安徽淮南232001 [2]合肥综合性国家科学中心能源研究院(安徽省能源实验室),安徽合肥230031

出　　处：《长江信息通信》2024年第10期1-4,共4页Changjiang Information & Communications

基　　金：安徽高校协同创新项目(GXXT-2022-004);合肥综合性国家科学中心能源研究院(安徽省能源实验室)项目(21KZS202、21KZS208)。

摘　　要：针对康普顿相机在稀疏投影数据成像情况下可能出现伪影等问题,文章设计了一种改进的U-net卷积神经网络,改善图像质量,该网络主要包含两个模块:解码层联合重构模块和高低频显示增强模块。首先在解码层阶段每一层引入高低频显示增强模块,该模块将注意力引入图像频域从而有效增强特征信息的表达能力;再利用解码层联合重构模块将不同解码层特征信息融合,减少图像重构过程中的特征稀释。实验表明改进后的模型比基础模型在SSIM和PSNR两个指标分别最大提升了0.05和2.651 dB。Aiming at the possible artifacts of Compton camera in the case of sparse projection data imaging,the article designs an improved U-nct convolutional ncural nctwork,which consists of two modules:the joint rcconstruction module of the decoding layer and the high and low fre-quency display enhancement module.Firstly,the high and low frequency display enhancement module is introduced in cach layer of the decoding laycr,which introduces the attention into the frequency domain of the image to effectively enhance the expression ability of the feature infor-mation;and then the joint reconstruction module of the decoding layer is used to fuse the feature information of different decoding layers to reduce the feature dilution in the process of image reconstruction.The experiments show that the improved model achieves a maximum improve-ment of 0.05 and 2.651 dB over the base model in the SSIM and PSNR metrics,respectively.

关 键 词：康普顿相机 U-net 通道注意力 傅里叶变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]