| 149 | 3 | 348 |
| 下载次数 | 被引频次 | 阅读次数 |
为了避免复杂背景对精细目标分类的影响,充分利用卷积神经网络提取的目标全局和局部信息进行细粒度任务的研究,提出了一种多尺度区域特征的细粒度目标检测与分类算法.该方法先使用FASTER-RCNN框架训练3个尺度区域的卷积模型进行多尺度目标区域定位,对定位的结果进行包围盒约束和海伦约束以优化提高定位的精确度,然后将提取多个尺度区域的特征进行组合,并用支持向量机训练细粒度分类器.在Caltech-UCSD鸟类数据集和Comp Cars车型数据集上进行实验测试.实验结果表明该算法在Caltech-UCSD鸟类数据集的分类正确率达到82. 8%,比没有使用多尺度区域特征的分类算法提高了7. 5%,比基于部件的分类方法提高了8. 9%;在Comp Cars车型数据集的分类正确率达到93. 5%,比没有使用多尺度区域特征的分类算法提高了8. 3%,比最优的Google Net精细目标分类算法提高了2. 3%,验证了该算法的有效性.
Abstract：Intending to reduce the influence of complex background on fine-grained classification,as well as to study the global information and local information of the target objects extracted from the convolutional neural network for fine-grained tasks,a fine-grained classification method based on multi-scale region feature was proposed. The method FASTER-RCNN framework was to train three convolution models to locate multi-scale object regions. Then the bounding box constraint and Helen constraint were applied to improve the location accuracy of the detected object. Finally,the extracted multi-scaled region features were combined to train a SVM classifier for fine-grained classification. The proposed method was tested in Caltech-UCSD bird datasets and CompCars vehicle datasets. The results showed that the accuracy of classification in Caltech-UCSD bird datasets was 82. 8%. It increased by 7. 5 % than the method without multi-scale region features. Compared with part-based RCNN,it increased by 8. 9 %. The results showed that the accuracy of classification in CompCars was 93. 5%. It increased by 8. 3 % than the method without multi-scale region features. Compared with GoogleNet,it increased by 2. 3 %.
[1] GOERING C,RODNER E,FREYTAG A,et al. Nonparametric part transfer for fine-grained recognition[C]∥Computer Vision and Pattern Recognition. Washington,2014:2489-2496.
[2] BERG T,BELHUMEUR P N. POOF:Part-based one-vs.-one features for fine-grained categorization,face verification,and Attribute Estimation[C]∥Computer Vision and Pattern Recognition. Washington,2013:955-962.
[3] CHAI Y,LEMPITSKY V,ZISSERMAN A. Symbiotic segmentation and part localization for fine-grained categorization[C]∥IEEE International Conference on Computer Vision. Washington,2013:321-328.
[4] ZHANG N,FARRELL R,IANDOLA F,et al. Deformable part descriptors for fine-grained recognition and attribute prediction[C]∥IEEE International Conference on Computer Vision. Washington,2014:729-736.
[5] YANG L,LUO P,CHEN C L,et al. A large-scale car dataset for fine-grained categorization and verification[C]∥Computer Vision and Pattern Recognition.Washington,2015:3973-3981.
[6] SZEGEDY C,LIU W,JIA Y,et al. Going deeper with convolutions[C]∥Computer Vision and Pattern Recognition.Washington,2015:1-9.
[7] SOCHOR J,HEROUT A,HAVEL J. Box Cars:3D boxes as CNN input for improved fine-grained vehicle recognition[C]∥IEEE Conference on Computer Vision and Pattern Recognition. Washington,2016:3006-3015.
[8] LIN Y L,MORARIU V I,HSU W,et al. Jointly optimizing 3D model fitting and fine-grained classification[C]∥European Conference on Computer Vision. Berlin,2014:466-480.
[9] SERMANET P,FROME A,REAL E. Attention for fine-grained categorization[J]. Computer science,2014,10(1):224-300.
[10] XIAO T,XU Y,YANG K,et al. The application of two-level attention models in deep convolutional neural network for finegrained image classification[C]∥Computer Vision and Pattern Recognition. Washington,2015:842-850.
[11] GAVVES E,FERNANDO B,SNOEK C G M,et al. Fine-grained categorization by alignments[C]∥IEEE International Conference on Computer Vision. Washington,2013:1713-1720.
[12] BRANSON S,VAN H G,BELONGIE S,et al. Bird species categorization using pose normalized deep convolutional Nets[C]∥British Machine Vision Conference. Berlin,2014:87-101.
[13] SIMON M,RODNER E. Neural activation constellations:unsupervised part model discovery with convolutional networks[C]∥IEEE International Conference on Computer Vision. Washington,2016:1143-1151.
[14] KRAUSE J,JIN H,YANG J,et al. Fine-grained recognition without part annotations[C]∥IEEE Conference on Computer Vision and Pattern Recognition.Washington,2015:5546-5555.
[15] GIRSHICK R,DONAHUE J,DARRELL T,et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]∥IEEE Conference on Computer Vision and Pattern Recognition. Washington,2014:580-587.
[16] HE K,ZHANG X,REN S,et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE transactions on pattern analysis and machine intelligence,2015,37(9):1904.
[17] GIRSHICK R. Fast R-CNN[C]∥IEEE International Conference on Computer Vision. Washington,2015:1440-1448.
[18] REN S,HE K,GIRSHICK R,et al. Fast R-CNN:towards real-time object detection with region proposal networks[J]. IEEE transactions on pattern analysis and machine intelligence,2015,39(6):1137-1149.
[19] REDMON J,DIVVALA S,GIRSHICK R,et al. You only look once:unified,real-time object detection[C]∥IEEE Conference on Computer Vision and Pattern Recognition. Washington,2016:779-788.
[20] LIU W,ANGUELOV D,ERHAN D,et al. SSD:single shot multibox detector[C]∥European Conference on Computer Vision.Berlin,2016:21-37.
[21] SUN X,WU P,HOI S C H. Face detection using deep learning:an improved faster R-CNN approach[J]. Neurocomputing,2018,299:42-50.
[22] ZHANG N,DONAHUE J,GIRSHICK R,et al. Part-based R-CNNs for fine-grained category detection[M].Cham:Springer Intermational Publishing,2014.
[23]唐斯琪,陶蔚,张梁梁,等.一种多列特征图融合的深度人群计数算法[J].郑州大学学报(理学版),2018,50(2):69-75.
[24]朱威威,赵岩松,李艳灵.一种基于集合划分的鲁棒性自适应模糊聚类分割算法[J].信阳师范学院学报(自然科学版),2019,32(1):146-152.
基本信息:
DOI：10.13705/j.issn.1671-6841.2018157
中图分类号:TP391.41;TP18
引用信息:
[1]熊昌镇,蒋杰.多尺度区域特征的细粒度分类算法研究[J],2019,51(03):55-60.DOI:10.13705/j.issn.1671-6841.2018157.
基金信息:
国家重点研发计划项目(2017YFC0821102)
2018-05-22
2018
2019-03-07
2019
2