基于S-PNet的触觉电阻抗成像后处理算法研究  

Research on post-processing algorithm of tactile electrical impedance tomography based on S-PNet

作　　者：袁晶晶 戎舟 Yuan Jingjing;Rong Zhou(College of Automation&College of Artificial Intelligence,NanJing University of Posts and Telecommunications,Nanjing 210023,China)

机构地区：[1]南京邮电大学自动化学院、人工智能学院,南京210023

出　　处：《国外电子测量技术》2024年第6期68-75,共8页Foreign Electronic Measurement Technology

摘　　要：电阻抗成像技术(EIT)因其非侵入式的特性为机器人柔性触觉传感器的压力点分布可视化提供了一种可行的方法。然而EIT逆问题具有高度的非线性和病态性,当多压力点相近时,重建图像的伪影会导致压力点间存在粘连。为解决上述问题,提出一种由特征提取、特征重建以及加强特征提取3个模块构成的S-PNet电阻抗成像后处理算法,实现对粘连压力点的分割以及形状重建。该算法使用金字塔池化结构加强特征提取,在增加极小计算量的情况下,能够提取到区分相近压力点边界的多尺度特征。采用均方根误差(RMSE)和结构相似度(SSIM)来评价后处理图像质量,实验得出RMSE的平均值为0.02,SSIM的平均值为0.97。仿真与实测结果均表明,与现有算法相比,基于S-PNet的后处理算法能够得到边界清晰且形状准确的结果。Electrical impedance tomography(EIT)provides a feasible method for visualizing the distribution of pressure points in flexible tactile sensors of robots due to its non-invasive characteristics.However,the inverse problem of EIT is highly nonlinear and ill-posed.When multiple pressure points are close to each other,the artifacts of the reconstructed image will lead to adhesion between pressure points.To solve the above problems,a post-processing algorithm named SPNet for EIT is proposed,which is composed of three modules:Feature extraction,feature reconstruction and enhanced feature extraction,to achieve the segmentation and shape reconstruction of adhesive pressure points.The algorithm uses a pyramid pooling structure to enhance feature extraction and can extract multi-scale features to distinguish the boundaries of close pressure points with minimal additional calculation.The post-processing image quality is evaluated by root mean square error(RMSE)and structural similarity(SSIM).The average value of RMSE is 0.02 and the average value of SSIM is 0.97.Both simulation and measured results show that compared with the existing algorithms,the postprocessing algorithm based on S-PNet can obtain images with clear boundaries and accurate shapes.

关 键 词：电阻抗成像 逆问题 图像后处理 深度学习 

分 类 号：R318[医药卫生—生物医学工程] TN911.73[医药卫生—基础医学]