·786· 智能系统学报 第17卷 4结束语 ageNet classification with deep convolutional neural net- works[J].Communications of the acm,2017,60(6): 本文提出了一种基于双层残差网络特征提 8490. 取、跳跃特征融合的图像语义分割网铬(DResnet), [6]SIMONYAN K,ZISSERMAN A.Very deep convolu- 它能够解决由特征提取网络带来的分辨率减小以 tional networks for large-scale image recognition[C]// 及空间细节信息损失问题。DResnet网络构建双 3rd International Conference on Learning Representa- 层残差特征提取网络与跳跃特征融合网络，双层 tions,ICLR 2015-Conference Track Proceedings.San 残差网络对标准训练集的图像同时进行两次特征 Diego:[s.n.]2015 提取操作，以获得更全面的图像细节特征信息； [7]SZEGEDY C,LIU Wei,JIA Yangqing,et al.Going deep- 跳跃特征融合网络在第1层与第2层开始融合跳 er with convolutions[Cl//2015 IEEE Conference on Com- 跃连接，更好地融合图像的背景信息与语义信 puter Vision and Pattern Recognition.Boston:IEEE. 息，使网络获得更高的特征提取能力与识别精 2015:1-9. 度。通过交通场景数据集CamVid和City- [8] TRONG V H,HYUN Y G,YOUNG K J,et al.Yielding scapes的实验结果表明，本文的DResnet网络模型 multi-fold training strategy for image classification of im- 在CamVid 12类上的评价指标PMoU、PMA、PAcc balanced weeds[J].Applied sciences,2021,11(8):3331. [9] MULLER D.KRAMER F.MIScnn:a framework for 分别提高了9.72%、3.63%、7.68%，在Cityscapes medical image segmentation with convolutional neural 20类上的评价指标PMIOU、PMPA、PAcc分别提高 networks and deep learning[EB/OL].(2019-10-21) 了3.42%、1.64%、3.48%。 [2021-06-13].https://www.semanticscholar.org/paper/MI 参考文献： Scnn%3A-A-Framework-for-Medical-Image-Segmenta- tion-M%C3%BCller-Kramer. [1]朱磊，滕奇志，龚剑.基于改进模糊C均值聚类和区域 [10]SHELHAMER E,LONG J,DARRELL T.Fully convo- 合并的矿物颗粒分割方法).科学技术与工程，2020， lutional networks for semantic segmentation[Cl//IEEE 20(34):14138-14145. transactions on pattern analysis and machine intelli- ZHU Lei.TENG Qizhi,GONG Jian.Mineral particle seg- gence.New York:IEEE,2015:640-651. mentation algorithm based on improved fuzzy C-means [11]EVERINGHAM M.ESLAMI S M A.GOOL L.et al and region merging[J].Science technology and engineer- The pascal visual object classes challenge:a retrospect- ing.2020.20(34):14138-14145. ive[J].International journal of computer vision,2015, [2]许林，孟娜，袁静，等.基于边缘检测和图像分割的超声 111(1:98-136. 诊断机器人控制系统设计[).计算机测量与控制， [12]SILBERMAN N.HOIEM D,KOHLI P,et al.Indoor 2020,28(8):125-129 segmentation and support inference from RGBD images XU Lin,MENG Na,YUAN Jing,et al.Design of ultra- [C]//European Conference on Computer Vision.Berlin: sonic diagnosis robot control system based on edge detec- Springer,2012:746-760. tion and image segmentation[J].Computer measurement [13]RONNEBERGER O.FISCHER P,BROX T.U-net &control,.2020,28(8):125-129. convolutional networks for biomedical image segmenta- [3]陈飞.改进的交互式Osu红外图像分割算法).计算 tion[M]//Lecture Notes in Computer Science.Cham: 机测量与控制.2020.28(9：248-251 Springer International Publishing,2015:234-241 CHEN Fei.An improved interactive otsu infrared image [14]XIAO Zhitao,LIU Bowen,GENG Lei,et al.Segmenta- segmentation algorithm[J].Computer measurement tion of lung nodules using improved 3D-UNet neural control,2020,28(9):248-251 network[J].Symmetry,2020,12(11):1787. [4]杨金鑫，杨辉华，李灵巧，等.结合卷积神经网络和超像 [15]GUAN Haixing,LI Hongliang,LI Rongqiang,et al. 素聚类的细胞图像分割方法[J].计算机应用研究， Face detection of innovation base based on faster 2018,35(5)1569-1572,1577 RCNN[M]//2021 International Conference on Applica- YANG Jinxin,YANG Huihua,LI Lingqiao,et al.Cell tions and Techniques in Cyber Intelligence.Cham: image segmentation method based on convolution neural Springer International Publishing,2021:158-165. network and super pixel clustering[]].Application re- [16]HE Kaiming,ZHANG Xiangyu,REN Shaoqing,et al. search of computers,2018,35(5):1569-1572,1577. Delving deep into rectifiers:surpassing human-level per- [5]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.Im- formance on ImageNet classification[C]//2015 IEEE In-4 结束语 本文提出了一种基于双层残差网络特征提 取、跳跃特征融合的图像语义分割网络（DResnet）。 它能够解决由特征提取网络带来的分辨率减小以 及空间细节信息损失问题。DResnet 网络构建双 层残差特征提取网络与跳跃特征融合网络，双层 残差网络对标准训练集的图像同时进行两次特征 提取操作，以获得更全面的图像细节特征信息； 跳跃特征融合网络在第 1 层与第 2 层开始融合跳 跃连接，更好地融合图像的背景信息与语义信 息，使网络获得更高的特征提取能力与识别精 度。通过交通场景数据 集 CamVi d 和 City - scapes 的实验结果表明，本文的 DResnet 网络模型 在 CamVid 12 类上的评价指标 PMIOU 、PMPA 、PACC 分别提高了 9.72%、3.63%、7.68%，在 Cityscapes 20 类上的评价指标 PMIOU 、PMPA 、PACC 分别提高 了 3.42%、1.64%、3.48%。 参考文献： 朱磊, 滕奇志, 龚剑. 基于改进模糊 C 均值聚类和区域 合并的矿物颗粒分割方法 [J]. 科学技术与工程, 2020, 20(34): 14138–14145. ZHU Lei, TENG Qizhi, GONG Jian. Mineral particle seg￾mentation algorithm based on improved fuzzy C-means and region merging[J]. Science technology and engineer￾ing, 2020, 20(34): 14138–14145. [1] 许林, 孟娜, 袁静, 等. 基于边缘检测和图像分割的超声 诊断机器人控制系统设计 [J]. 计算机测量与控制, 2020, 28(8): 125–129. XU Lin, MENG Na, YUAN Jing, et al. Design of ultra￾sonic diagnosis robot control system based on edge detec￾tion and image segmentation[J]. Computer measurement & control, 2020, 28(8): 125–129. [2] 陈飞. 改进的交互式 Otsu 红外图像分割算法 [J]. 计算 机测量与控制, 2020, 28(9): 248–251. CHEN Fei. An improved interactive otsu infrared image segmentation algorithm[J]. Computer measurement & control, 2020, 28(9): 248–251. [3] 杨金鑫, 杨辉华, 李灵巧, 等. 结合卷积神经网络和超像 素聚类的细胞图像分割方法 [J]. 计算机应用研究, 2018, 35(5): 1569–1572,1577. YANG Jinxin, YANG Huihua, LI Lingqiao, et al. Cell image segmentation method based on convolution neural network and super pixel clustering[J]. Application re￾search of computers, 2018, 35(5): 1569–1572,1577. [4] [5] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. Im￾ageNet classification with deep convolutional neural net￾works[J]. Communications of the acm, 2017, 60(6): 84–90. SIMONYAN K, ZISSERMAN A. Very deep convolu￾tional networks for large-scale image recognition[C]// 3rd International Conference on Learning Representa￾tions, ICLR 2015-Conference Track Proceedings. San Diego: [s. n. ], 2015. [6] SZEGEDY C, LIU Wei, JIA Yangqing, et al. Going deep￾er with convolutions[C]//2015 IEEE Conference on Com￾puter Vision and Pattern Recognition. Boston: IEEE, 2015: 1−9. [7] TRONG V H, HYUN Y G, YOUNG K J, et al. Yielding multi-fold training strategy for image classification of im￾balanced weeds[J]. Applied sciences, 2021, 11(8): 3331. [8] MÜLLER D, KRAMER F. MIScnn: a framework for medical image segmentation with convolutional neural networks and deep learning[EB/OL]. (2019−10−21) [2021−06−13].https://www.semanticscholar.org/paper/MI Scnn%3A-A-Framework-for-Medical-Image-Segmenta￾tion-M%C3%BCller-Kramer. [9] SHELHAMER E, LONG J, DARRELL T. Fully convo￾lutional networks for semantic segmentation[C]//IEEE transactions on pattern analysis and machine intelli￾gence. New York: IEEE, 2015 : 640−651. [10] EVERINGHAM M, ESLAMI S M A, GOOL L, et al. The pascal visual object classes challenge: a retrospect￾ive[J]. International journal of computer vision, 2015, 111(1): 98–136. [11] SILBERMAN N, HOIEM D, KOHLI P, et al. Indoor segmentation and support inference from RGBD images [C]//European Conference on Computer Vision. Berlin: Springer, 2012: 746−760. [12] RONNEBERGER O, FISCHER P, BROX T. U-net: convolutional networks for biomedical image segmenta￾tion[M]//Lecture Notes in Computer Science. Cham: Springer International Publishing, 2015: 234−241. [13] XIAO Zhitao, LIU Bowen, GENG Lei, et al. Segmenta￾tion of lung nodules using improved 3D-UNet neural network[J]. Symmetry, 2020, 12(11): 1787. [14] GUAN Haixing, LI Hongliang, LI Rongqiang, et al. Face detection of innovation base based on faster RCNN[M]//2021 International Conference on Applica￾tions and Techniques in Cyber Intelligence. Cham: Springer International Publishing, 2021: 158−165. [15] HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Delving deep into rectifiers: surpassing human-level per￾formance on ImageNet classification[C]//2015 IEEE In- [16] ·786· 智 能 系 统 学 报 第 17 卷