基于深度卷积神经网络的三维模型检索
|
摘要
为了更有效地利用三维模型数据集进行特征的自主学习,提出一种使用自然图像作为输入源,以三维模型的较优视图集为基础,通过深度卷积神经网络的训练提取深度特征用于检索的三维模型检索方法。首先,从多个视点对三维模型进行视图提取,并根据灰度熵的排序选取较优视图;然后,通过深度卷积神经网络对视图集进行训练,从而提取较优视图的深度特征并进行降维,同时,对输入的自然图像提取边缘轮廓图,经过相似度匹配获得一组三维模型;最后,基于检索结果中同类模型总数占检索列表长度的比例对列表进行重排序,从而获得最终的检索结果。实验结果表明,该算法能够有效利用深度卷积神经网络对三维模型的视图进行深度特征提取,同时降低了输入源的获取难度,有效提高了检索效果。
In order to make use of 3D model data set for feature self-learning more effectrively,this paper proposed a 3D model retrieval method,in which the natural images are as input sources,a better view set of 3D model is as basis,and the depth features obtained through training deep convolutional neural networks are applied for retrieval.Firstly,the 3D model views are extracted from multiple viewpoints and the optimal view is selected according to the order of gray entropy.Secondly,the view set is trained through deep convolution neural network,the depth features of the optimal view are extracted and its dimension is reduced.At the same time,edge contouring is extracted for the input natural image and a set of 3D models is gotten after similarity matching.Finally,according to the ratio of the total number of similar models to the length of the search list in the retrieval result,the retrieval list is reordered and the final result will be gained.Experimental results show that the algorithm can effectively use depth convolution neural network to extract the depth feature of the view of the 3Dmodel,meanwhile reduce the difficulty of obtaining the input source,and improve the retrieval efficiency effectively.
In order to make use of 3D model data set for feature self-learning more effectrively,this paper proposed a 3D model retrieval method,in which the natural images are as input sources,a better view set of 3D model is as basis,and the depth features obtained through training deep convolutional neural networks are applied for retrieval.Firstly,the 3D model views are extracted from multiple viewpoints and the optimal view is selected according to the order of gray entropy.Secondly,the view set is trained through deep convolution neural network,the depth features of the optimal view are extracted and its dimension is reduced.At the same time,edge contouring is extracted for the input natural image and a set of 3D models is gotten after similarity matching.Finally,according to the ratio of the total number of similar models to the length of the search list in the retrieval result,the retrieval list is reordered and the final result will be gained.Experimental results show that the algorithm can effectively use depth convolution neural network to extract the depth feature of the view of the 3Dmodel,meanwhile reduce the difficulty of obtaining the input source,and improve the retrieval efficiency effectively.
引文
[1]SCHMIDHUBER J.Deep learning in neural networks:an overview[J].Neural Networks the Official Journal of the International Neural Network Society,2014,61(1):85-117.
[2]SUN Z Y,LU C X,SHI Z Z,et al.Research and Progress of Deep Learning[J].Computer Science,2016,43(2):1-8.(in Chinese)孙志远,鲁成祥,史忠植,等.深度学习研究与进展[J].计算机科学,2016,43(2):1-8.
[3]LUO W,SCHWING A G,URTASUN R.Efficient Deep Learning for Stereo Matching[C]∥IEEE Conference on Computer Vision and Pattern Recognition.IEEE Computer Society,2016:5695-5703.
[4]ZHANG J,BAI C,NEZAN J F,et al.Joint motion model for local stereo video-matching method[J].Optical Engineering,2015,54(12):123108.
[5]SHI M,XIE F,ZI Y,et al.Cloud detection of remote sensing images by deep learning[C]∥Geoscience and Remote Sensing Symposium.IEEE,2016:701-704.
[6]ZHANG J,SHANG J,ZHANG G.Verification for Different Contrail Parameterizations Based on Integrated Satellite Observation and ECMWF Reanalysis Data[J].Advances in Meteorology,2017,2017(1):1-11.
[7]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]∥International Conference on Neural Information Processing Systems.Curran Associates Inc.,2012:1097-1105.
[8]SIMONYAN K,ZISSERMAN A.Very Deep Convolutional Networks for Large-Scale Image Recognition[EB/OL].https://cn.arXiv.org/abs/1409.1556.2014.
[9]ALZU’BI A,AMIRA A,RAMZAN N.Content Based Image Retrieval with Compact Deep Convolutional Features[J].Neurocomputing,2017,249(2):95-105.
[10]SALVADOR A,GIROINIETO X,MARQUES F,et al.Faster R-CNN Features for Instance Search[C]∥IEEE Conference on Computer Vision and Pattern Recognition Workshops.IEEEComputer Society,2016:394-401.
[11]LIU Y,PAN Y,XIA R K,et al.FP-CNNH:A Fast Image Hashing Algorithm Based on Depth Convolution Neural Network[J].Computer Science,2016,43(9):39-46.(in Chinese)刘冶,潘炎,夏榕楷,等.FP-CNNH:一种基于深度卷积神经网络的快速图像哈希算法[J].计算机科学,2016,43(9):39-46.
[12]ZHU Z,WANG X,BAI S,et al.Deep Learning Representation using Autoencoder for 3DShape Retrieval[J].Neurocomputing,2016,204(C):41-50.
[13]LI X X,CAO Q,WEI S.3Dobject retrieval based on multiview convolutional neural networks[J].Multimedia Tools&Applications,2017,76(19):20111-201214.
[14]WANG F,KANG L,LI Y.Sketch-based 3dshape retrieval using convolutional neural networks[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2015:1875-1883.
[15]DECARLO D,FINKELSTEIN A,RUSINKIEWICZ S,et al.Suggestive contours for conveying shape[C]∥ACM Siggraph.ACM,2003:848-855.
[16]DING L,GOSHTASBY A.On the Canny edge detector[J].Pattern Recognition,2001,34(3):721-725.
[17]SANG M Y,SCHERER M,SCHRECK T,et al.Sketch-based3D model retrieval using diffusion tensor fields of suggestive contours[C]∥International Conference on Multimedea 2010.Firenze,Italy,DBLP,2010:193-200.
[18]LI B,LU Y,JOHAN H.Sketch-based 3D model retrieval by viewpoint entropy based adaptive view clustering[C]∥Eurographics Workshop on 3dObject Retrieval.Eurographics Association,2013:49-56.
[19]HONG Y,KIM J,HONG Y,et al.A 2D-View Depth Image and CNN-Based 3D Model Identification Method[J].Applied Sciences,2017,7(10):988-1001.
[20]CLEVERT D,UNTERTHINER T,HOCHREITER S.Fast and Accurate Deep Network Learning by Exponential Linear Units(ELUs)[EB/OL].https://cnarXiv.org/abs/1511.07289.2015.
[21]LI B,LU Y,LI C,et al.A comparison of 3Dshape retrieval methods based on a large-scale benchmark supporting multimodal queries[J].Computer Vision and Image Understanding,2015,131(1):1-27.
[22]LI B,LU Y,LI C,et al.Shrec’14track:extended large scale sketch-based 3Dshape retrieval[C]∥Eurographics Workshop on 3DObject Retrieval.2014.
[23]LIU Z,YIN S C,PAN X,et al.3Dmodel retrieval based on feature lines[J].Journal of Computer-Aided Design&Computer Graphics,2016,28(9):1512.(in Chinese)刘志,尹世超,潘翔,等.基于特征线条的三维模型检索方法[J].计算机辅助设计与图形学学报,2016,28(9):1512.
[24]HUA S,JIANG Q,ZHONG Q.3D Model Retrieval Based on Multi-View SIFT Feature[M]∥Communication Systems and Information Technology.Springer Berlin Heidelberg,2011:163-169.
[2]SUN Z Y,LU C X,SHI Z Z,et al.Research and Progress of Deep Learning[J].Computer Science,2016,43(2):1-8.(in Chinese)孙志远,鲁成祥,史忠植,等.深度学习研究与进展[J].计算机科学,2016,43(2):1-8.
[3]LUO W,SCHWING A G,URTASUN R.Efficient Deep Learning for Stereo Matching[C]∥IEEE Conference on Computer Vision and Pattern Recognition.IEEE Computer Society,2016:5695-5703.
[4]ZHANG J,BAI C,NEZAN J F,et al.Joint motion model for local stereo video-matching method[J].Optical Engineering,2015,54(12):123108.
[5]SHI M,XIE F,ZI Y,et al.Cloud detection of remote sensing images by deep learning[C]∥Geoscience and Remote Sensing Symposium.IEEE,2016:701-704.
[6]ZHANG J,SHANG J,ZHANG G.Verification for Different Contrail Parameterizations Based on Integrated Satellite Observation and ECMWF Reanalysis Data[J].Advances in Meteorology,2017,2017(1):1-11.
[7]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]∥International Conference on Neural Information Processing Systems.Curran Associates Inc.,2012:1097-1105.
[8]SIMONYAN K,ZISSERMAN A.Very Deep Convolutional Networks for Large-Scale Image Recognition[EB/OL].https://cn.arXiv.org/abs/1409.1556.2014.
[9]ALZU’BI A,AMIRA A,RAMZAN N.Content Based Image Retrieval with Compact Deep Convolutional Features[J].Neurocomputing,2017,249(2):95-105.
[10]SALVADOR A,GIROINIETO X,MARQUES F,et al.Faster R-CNN Features for Instance Search[C]∥IEEE Conference on Computer Vision and Pattern Recognition Workshops.IEEEComputer Society,2016:394-401.
[11]LIU Y,PAN Y,XIA R K,et al.FP-CNNH:A Fast Image Hashing Algorithm Based on Depth Convolution Neural Network[J].Computer Science,2016,43(9):39-46.(in Chinese)刘冶,潘炎,夏榕楷,等.FP-CNNH:一种基于深度卷积神经网络的快速图像哈希算法[J].计算机科学,2016,43(9):39-46.
[12]ZHU Z,WANG X,BAI S,et al.Deep Learning Representation using Autoencoder for 3DShape Retrieval[J].Neurocomputing,2016,204(C):41-50.
[13]LI X X,CAO Q,WEI S.3Dobject retrieval based on multiview convolutional neural networks[J].Multimedia Tools&Applications,2017,76(19):20111-201214.
[14]WANG F,KANG L,LI Y.Sketch-based 3dshape retrieval using convolutional neural networks[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2015:1875-1883.
[15]DECARLO D,FINKELSTEIN A,RUSINKIEWICZ S,et al.Suggestive contours for conveying shape[C]∥ACM Siggraph.ACM,2003:848-855.
[16]DING L,GOSHTASBY A.On the Canny edge detector[J].Pattern Recognition,2001,34(3):721-725.
[17]SANG M Y,SCHERER M,SCHRECK T,et al.Sketch-based3D model retrieval using diffusion tensor fields of suggestive contours[C]∥International Conference on Multimedea 2010.Firenze,Italy,DBLP,2010:193-200.
[18]LI B,LU Y,JOHAN H.Sketch-based 3D model retrieval by viewpoint entropy based adaptive view clustering[C]∥Eurographics Workshop on 3dObject Retrieval.Eurographics Association,2013:49-56.
[19]HONG Y,KIM J,HONG Y,et al.A 2D-View Depth Image and CNN-Based 3D Model Identification Method[J].Applied Sciences,2017,7(10):988-1001.
[20]CLEVERT D,UNTERTHINER T,HOCHREITER S.Fast and Accurate Deep Network Learning by Exponential Linear Units(ELUs)[EB/OL].https://cnarXiv.org/abs/1511.07289.2015.
[21]LI B,LU Y,LI C,et al.A comparison of 3Dshape retrieval methods based on a large-scale benchmark supporting multimodal queries[J].Computer Vision and Image Understanding,2015,131(1):1-27.
[22]LI B,LU Y,LI C,et al.Shrec’14track:extended large scale sketch-based 3Dshape retrieval[C]∥Eurographics Workshop on 3DObject Retrieval.2014.
[23]LIU Z,YIN S C,PAN X,et al.3Dmodel retrieval based on feature lines[J].Journal of Computer-Aided Design&Computer Graphics,2016,28(9):1512.(in Chinese)刘志,尹世超,潘翔,等.基于特征线条的三维模型检索方法[J].计算机辅助设计与图形学学报,2016,28(9):1512.
[24]HUA S,JIANG Q,ZHONG Q.3D Model Retrieval Based on Multi-View SIFT Feature[M]∥Communication Systems and Information Technology.Springer Berlin Heidelberg,2011:163-169.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |