第2期 闫涵，等：多感知兴趣区域特征融合的图像识别方法 ·269· 区域及学习到的特征有所差异。针对上述问题构 automatica sinica,2014,40(10):2346-2355. 建了多个模型融合机制，通过借鉴DenseNet模型 [10]YE Qihong,XIANG Ming,CUI Zhendong.Fingerprint 的多尺度特征拼接及ResNet特征相加机制，设计 image enhancement algorithm based on two dimension 了Multi view Fusion模型、Multi view Fusion EMD and Gabor filter[J].Procedia engineering,2012,29: tiny模型和Voted Model。实验结果表明本文算法 1840-1844. 在相似目标的二分类问题上具有更高的识别准确 [11]张丽琼，王炳和.基于小波变换的脉象信号特征提取方 率。下一步的研究方向可放在网络结构轻量化与 法).数据采集与处理，2004,19(3：323-328. 模型的加速上。 ZHANG Liqiong,WANG Binghe.Feature extraction methods for pulse signal based on wavelet transform[J]. 参考文献： Journal of data acquisition processing,2004,19(3): [1]SPENCER JR B F.HOSKERE V.NARAZAKI Y.Ad- 323-328. vances in computer vision-based civil infrastructure in- [12]YAGHOOBI H.MANSOURI H.FARSANGI MA E.et al spection and monitoring[J].Engineering,2019,5(2): Determining the fragmented rock size distribution using 199-222. textural feature extraction of images[J].Powder techno- [2]CHELLAPPA R.The changing fortunes of pattern recog- l1ogy,2019,342:630-641. nition and computer vision[J].Image and vision comput- [13]MESNIL G,BORDES A,WESTON J,et al.Learning se- ing,2016,55:3-5. mantic representations of objects and their parts[J].Ma- [3]雷明.机器学习与应用[M.北京：清华大学出版社， chine learning,2014,94(2):281-301. 2019:26-33 [14]胡越，罗东阳，花奎，等，关于深度学习的综述与讨 [4]MARY N A B,DHARMA D.Coral reef image classifica- 论[U.智能系统学报，2019,14(1)：1-19 tion employing improved LDP for feature extraction[J]. HU Yue,LUO Dongyang,HUA Kui,et al.Overview on Journal of visual communication and image representation, deep learning[J].CAAI transactions on intelligent sys- 2017.49:225-242. tems,2019,141):1-19 [5]YU Hua,YANG Jie.A direct LDA algorithm for high-di- [15]SERMANET P,EIGEN D,ZHANG Xiang,et al.Over- mensional data-with application to face recognition[J]. feat:integrated recognition,localization and detection us- Pattern recognition,2001,34(10):2067-2070. ing convolutional networks[J].Computer science,2013. [6]刘丽，匡纲要.图像纹理特征提取方法综述)中国图象 [16]ZEILER M D.KRISHNAN D.TAYLOR G W.et al.De- 图形学报，2009,14(4)：622-635 convolutional networks[C]//Proceedings of 2010 IEEE LIU Li,KUANG Gangyao.Overview of image textural Computer Society Conference on Computer Vision and feature extraction methods[J].Journal of image and graph- Pattern Recognition.San Francisco,USA,2010: ics.2009.14(4):622-635. 2528-2535. [7]李磊，董卓丽.利用改进图割的彩色图像分割算法】.武 [17]KRIZHEVSKY A.SUTSKEVER I.HINTON G E.Im- 汉大学学报·信息科学版，2014,39(12)：1504-1508. ageNet classification with deep convolutional neural net- LI Lei.DONG Zhuoli.Color image segmentation using works[J].Communications of the ACM,2017,60(6) improved graph cuts[J].Geomatics and Information Sci- 84-90. ence of Wuhan University,2014,39(12):1504-1508 [18]SIMONYAN K.ZISSERMAN A.Very deep convolu- [8]弓晓虹，郑音飞，秦佳乐，等.基于乘性梯度的医学超声 tional networks for large-scale image recognition[EB/OL]. 图像边缘检测算法U.浙江大学学报（工学版），2014， https://arxiv.org/abs/1409.1556,2014. 48(10):1871-1878 [19]HE Kaiming,ZHANG Xiangyu,REN Shaoqing,et al. GONG Xiaohong,ZHENG Yinfei,QIN Jiale,et al.Multi- Deep residual learning for image recognition[C]//Proceed- plicative gradient based edge detection method for medic- ings of 2016 IEEE Conference on Computer Vision and al ultrasound image[J].Journal of Zhejiang University(En- Pattern Recognition.Las Vegas,USA,2016:770-778. gineering Science),2014,48(10):1871-1878. [20]HUANG Gao,LIU Zhuang,VAN DER MAATEN L,et [9]张桂梅，张松，储珺.一种新的基于局部轮廓特征的目标 al.Densely connected convolutional networks[C//Pro- 检测方法U.自动化学报，2014,40(10)：2346-2355。 ceedings of 2017 IEEE Conference on Computer Vision ZHANG Guimei,ZHANG Song,CHU Jun.A new object and Pattern Recognition.Honolulu,USA,2017: detection algorithm using local contour features[].Acta 2261-2269.区域及学习到的特征有所差异。针对上述问题构 建了多个模型融合机制，通过借鉴 DenseNet 模型 的多尺度特征拼接及 ResNet 特征相加机制，设计 了 Multi view Fusion 模型、Multi view Fusion tiny 模型和 Voted Model。实验结果表明本文算法 在相似目标的二分类问题上具有更高的识别准确 率。下一步的研究方向可放在网络结构轻量化与 模型的加速上。 参考文献： SPENCER JR B F, HOSKERE V, NARAZAKI Y. Ad￾vances in computer vision-based civil infrastructure in￾spection and monitoring[J]. Engineering, 2019, 5(2): 199–222. [1] CHELLAPPA R. The changing fortunes of pattern recog￾nition and computer vision[J]. Image and vision comput￾ing, 2016, 55: 3–5. [2] 雷明. 机器学习与应用 [M]. 北京: 清华大学出版社, 2019: 26−33. [3] MARY N A B, DHARMA D. Coral reef image classifica￾tion employing improved LDP for feature extraction[J]. Journal of visual communication and image representation, 2017, 49: 225–242. [4] YU Hua, YANG Jie. A direct LDA algorithm for high-di￾mensional data—with application to face recognition[J]. Pattern recognition, 2001, 34(10): 2067–2070. [5] 刘丽, 匡纲要. 图像纹理特征提取方法综述 [J]. 中国图象 图形学报, 2009, 14(4): 622–635. LIU Li, KUANG Gangyao. Overview of image textural feature extraction methods[J]. Journal of image and graph￾ics, 2009, 14(4): 622–635. [6] 李磊, 董卓丽. 利用改进图割的彩色图像分割算法 [J]. 武 汉大学学报 • 信息科学版, 2014, 39(12): 1504–1508. LI Lei, DONG Zhuoli. Color image segmentation using improved graph cuts[J]. Geomatics and Information Sci￾ence of Wuhan University, 2014, 39(12): 1504–1508. [7] 弓晓虹, 郑音飞, 秦佳乐, 等. 基于乘性梯度的医学超声 图像边缘检测算法 [J]. 浙江大学学报(工学版), 2014, 48(10): 1871–1878. GONG Xiaohong, ZHENG Yinfei, QIN Jiale, et al. Multi￾plicative gradient based edge detection method for medic￾al ultrasound image[J]. Journal of Zhejiang University (En￾gineering Science), 2014, 48(10): 1871–1878. [8] 张桂梅, 张松, 储珺. 一种新的基于局部轮廓特征的目标 检测方法 [J]. 自动化学报, 2014, 40(10): 2346–2355. ZHANG Guimei, ZHANG Song, CHU Jun. A new object detection algorithm using local contour features[J]. Acta [9] automatica sinica, 2014, 40(10): 2346–2355. YE Qihong, XIANG Ming, CUI Zhendong. Fingerprint image enhancement algorithm based on two dimension EMD and Gabor filter[J]. Procedia engineering, 2012, 29: 1840–1844. [10] 张丽琼, 王炳和. 基于小波变换的脉象信号特征提取方 法 [J]. 数据采集与处理, 2004, 19(3): 323–328. ZHANG Liqiong, WANG Binghe. Feature extraction methods for pulse signal based on wavelet transform[J]. Journal of data acquisition & processing, 2004, 19(3): 323–328. [11] YAGHOOBI H, MANSOURI H, FARSANGI M A E, et al. Determining the fragmented rock size distribution using textural feature extraction of images[J]. Powder techno￾logy, 2019, 342: 630–641. [12] MESNIL G, BORDES A, WESTON J, et al. Learning se￾mantic representations of objects and their parts[J]. Ma￾chine learning, 2014, 94(2): 281–301. [13] 胡越, 罗东阳, 花奎, 等. 关于深度学习的综述与讨 论 [J]. 智能系统学报, 2019, 14(1): 1–19. HU Yue, LUO Dongyang, HUA Kui, et al. Overview on deep learning[J]. CAAI transactions on intelligent sys￾tems, 2019, 14(1): 1–19. [14] SERMANET P, EIGEN D, ZHANG Xiang, et al. Over￾feat: integrated recognition, localization and detection us￾ing convolutional networks[J]. Computer science, 2013. [15] ZEILER M D, KRISHNAN D, TAYLOR G W, et al. De￾convolutional networks[C]//Proceedings of 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. San Francisco, USA, 2010: 2528−2535. [16] 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. [17] SIMONYAN K, ZISSERMAN A. Very deep convolu￾tional networks for large-scale image recognition[EB/OL]. https: //arxiv.org/abs/1409.1556, 2014. [18] HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]//Proceed￾ings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, USA, 2016: 770−778. [19] HUANG Gao, LIU Zhuang, VAN DER MAATEN L, et al. Densely connected convolutional networks[C]//Pro￾ceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, USA, 2017: 2261−2269. [20] 第 2 期 闫涵，等：多感知兴趣区域特征融合的图像识别方法 ·269·