基于强监督部件模型的遥感图像目标检测

doi:10.11772/j.issn.1001-9081.2016.06.1714

计算机应用 ›› 2016, Vol. 36 ›› Issue (6): 1714-1718.DOI: 10.11772/j.issn.1001-9081.2016.06.1714

基于强监督部件模型的遥感图像目标检测

周福送^1,2, 霍宏^1,2, 万卫兵^1,2, 方涛^1,2

1. 上海交通大学自动化系, 上海 200240;
2. 系统控制与信息处理教育部重点实验室, 上海 200240

收稿日期:2015-11-27 修回日期:2016-01-27 出版日期:2016-06-10 发布日期:2016-06-08
通讯作者: 周福送
作者简介:周福送(1992-),男,江西九江人,硕士研究生,主要研究方向:模式识别、目标检测;霍宏(1972-),女,辽宁本溪人,讲师,博士,主要研究方向:模式识别、遥感图像理解;万卫兵(1969-),男,江西南昌人,讲师,博士,主要研究方向:视频图像处理、数据融合;方涛(1965-),男,四川彭山人,教授,博士,主要研究方向:遥感图像理解、图像处理、模式识别。
基金资助:
国家973计划项目(2012CB719903);国家自然科学基金委创新研究群体资助项目(61221003);国家自然科学基金青年科学基金资助项目(41101386)。

Object detection in remote sensing imagery based on strongly-supervised deformable part models

ZHOU Fusong^1,2, HUO Hong^1,2, WAN Weibing^1,2, FANG Tao^1,2

1. Department of Automation, Shanghai Jiao Tong University, Shanghai 200240, China;
2. Key Laboratory of System Control and Information Processing, Ministry of Education, Shanghai 200240, China

Received:2015-11-27 Revised:2016-01-27 Online:2016-06-10 Published:2016-06-08
Supported by:
This work is partially supported by the National Basic Research Program (973 Program) of China (2012CB719903), the Science Fund for Creative Research Groups of the National Natural Science Foundation of China (61221003), the Young Scientists Fund of the National Natural Science Foundation of China (41101386).

摘要/Abstract

摘要： 针对遥感图像中由于背景复杂、目标外观多样和方向任意而导致的检测精度不高的问题,提出一种基于强监督的部件模型方法。该方法针对目标的每个方向范围训练子模型,同时训练集除了标注出目标的外接矩形,还标注出每个部件的位置及其语义。模型训练时,首先,通过对训练集图像建立多尺度方向梯度直方图(HOG)特征金字塔,且根据目标部件标注信息采用最小生成树(MST)算法初始化模型结构;其次通过隐支持向量机(LSVM)方法训练出多个对应不同方向区域的子模型,每个子模型由一个目标滤波器和多个两倍分辨率的部件滤波器,以及位置关系模型组成,多个子模型最终合并成用来检测的混合模型。目标检测时,类似地建立多尺度特征金字塔,然后利用训练滤波器模型在特征金字塔上以滑动窗口的方式计算匹配响应得分,对响应得分设置阈值且采用非极大值抑制(NMS)算法来获得优化后的检测结果。该方法在自建的遥感数据集上目标检测精度达到了89.4%,对比弱监督部件模型(DPM)、分类器模板集成(Exemplar-SVMs)和方向梯度直方图-支持向量机(HOG-SVM)方法中的最高精度,所提方法提升了4个百分点。实验结果表明,所提算法能够在解决方向和背景复杂问题上有一定的提升,而且可以应用于机场军事飞机目标检测。

关键词: 目标检测, 遥感图像, 部件模型, 混合模型, 强监督

Abstract: The object detection of remote sensing imagery has lower detection accuracy caused by complexity of background, target appearance variety and arbitrary orientation. In order to solve the problem, a method based on strongly-supervised deformable part models was proposed. Then multiple sub-models in each direction range of the object were trained. In addition, the object bounding rectangle, position and semantic information of every part were labeled. In the model training stage, firstly, multi-scale Histogram of Oriented Gradients (HOG) feature pyramid for every training image was constructed, and the model structure was initialized according to object-part label information and Minimum Span Tree (MST). Secondly, the sub-models corresponding to every direction region were trained using Latent Support Vector Machine (LSVM). Every sub-model was consisted of a object filter, multiple twice resolution part filters, and the position relation model. Finally, the mixture model was merged from all sub-models to detect object. In the object detection stage, the multi-scale feature pyramid was also firstly constructed, then matching response score in feature pyramid was computed using mixture training filter model and sliding window. Optimized detection results could be obtained by setting threshold for the response score and adopting Non-Maximum Suppression (NMS) algorithm. The object detection accuracy of the proposed method is 89.4% on self-built remote sensing data sets, compared to the highest accuracy among weakly-supervised Deformable Part Model (DPM), Exemplar Support Vector Machines (Exemplar-SVMs) and Histogram of Oriented Gradients-Support Vector Machine (HOG-SVM), the proposed algorithm has an improvement about 4 percentage points in detection behavior. The experimental results show that the proposed algorithm could has some improvement in solving above mentioned direction and background complex problems, and also can be applied in object detection of the airport military airplane.

Key words: object detection, remote sensing imagery, part model, mixture model, strongly-supervised

中图分类号:

TP391.41

周福送, 霍宏, 万卫兵, 方涛. 基于强监督部件模型的遥感图像目标检测[J]. 计算机应用, 2016, 36(6): 1714-1718.

ZHOU Fusong, HUO Hong, WAN Weibing, FANG Tao. Object detection in remote sensing imagery based on strongly-supervised deformable part models[J]. Journal of Computer Applications, 2016, 36(6): 1714-1718.

参考文献

[1] 刘扬,付征叶,郑逢斌.高分辨率遥感影像目标分类与识别研究进展[J].地球信息科学学报,2015,17(9):1080-1091.(LIU Y, FU Z Y, ZHENG F B. Object classification and recognition research progress in high resolution remote image[J]. Journal of Geo-information Science, 2015, 17(9): 1080-1091.)
[2] DALAL N, TRIGGS B. Histograms of oriented gradients for human detection[C]//Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Washington, DC: IEEE Computer Society, 2005, 1: 886-893.
[3] 姚远,姜志国,张浩鹏.基于层次化分类器的遥感图像飞机目标检测[J].航天返回与遥感,2014,35(5):88-94.(YAO Y, JIANG Z G, ZHANG H P. Aircraft detection in remote sensing image based on hierarchical classifiers[J]. Spacecraft Recovery & Remote Sensing, 2014, 35(5): 88-94.)
[4] 张宇,何楚,石博,等.逐层特征选择的多层部件模型用于遥感图像飞机目标检测[J].武汉大学学报(信息科学版),2014,39(12):1406-1411.(ZHANG Y, HE C, SHI B, et al. Multi-layer feature selection based hierarchal component model for aero-plane detection on remote sensing image[J]. Geomatics and Information Science of Wuhan University, 2014, 39(12): 1406-1411.)
[5] FENSZWALB P F, GIRSHICK R B, MCALLETER D, et al. Object detection with discriminatively trained part-based models[J]. IEEE Transaction on Pattern Analysis and Machine Intelligence, 2010, 32(9): 1627-1645.
[6] 胡正平,杨建秀.HOG特征混合模型结合隐SVM的感兴趣目标检测定位算法[J].信号处理,2011,27(8):1206-1212.(HU Z P, YANG J X. Object localization algorithm based on mixture model of HOG feature and LSVM[J]. Signal Processing, 2011, 27(8): 1206-1212.)
[7] 何楚,张宇,廖紫纤,等.基于分层自适应部分模型的遥感图像飞机目标检测[J].武汉大学学报:信息科学版,2013,38(6):656-670.(HE C, ZHANG Y, LIAO Z Q, et al. Object detection of aero-plane based on hierarchical adaptive part-based model[J]. Geomatics and Information Science of Wuhan University, 2013, 38(6): 656-670.)
[8] LOVE D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004, 60(2): 91-110.
[9] PANDEY M, LAZEBNIK S. Scene recognition and weakly supervised object localization with deformable part-based models[C]//Proceedings of the 2011 International Conference on Computer Vision. Washington, DC: IEEE Computer Society, 2011: 1307-1314.
[10] AZIZPOUR H, LAPTEV I. Object detection using strongly-supervised deformable part models[C]//ECCV 2012: Proceedings of the 12th European Conference on Computer Vision, LNCS 7572. Berlin: Springer, 2012: 836-849.
[11] MALISIEWICZ T, GUPTA A, EFROS A A. Ensemble of exemplar-SVMs for object detection and beyond[C]//Proceedings of the 2011 International Conference on Computer Vision. Washington, DC: IEEE Computer Society, 2011: 89-96.
[12] PARK D, RAMANAN D, FOWLKES C. Multiresolution models for object detection[C]//ECCV 2010: Proceedings of the 11th European Conference on Computer Vision, LNCS 6314. Berlin: Springer, 2010: 241-254.
[13] LU S X, WANG Z. A comparison among four SVM classification methods: LSVM, NLSVM, SSVM and NSVM[C]//Proceedings of the 2004 International Conference on Machine Learning and Cybernetics. Piscataway, NJ: IEEE, 2004, 7: 4277-4282.
[14] FELZENSZWALB P F, HUTTENLOCHER D P. Distance transforms of sampled functions[J]. Theory of Computing, 2012, 8(19): 415-428.
[15] ANDREWS S, TSOCHANTARIDIS I, HOFMANN T. Support vector machines for multiple-instance learning[C]//Proceedings of the 2002 Advances in Neural Information Processing Systems. Cambridge: MIT Press, 2002: 561-568.
[16] CHENG G, HAN J, GUO L, et al. Object detection in remote sensing imagery using a discriminatively trained mixture model[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2013, 85: 32-43.
[17] FELZENSZWALB P F, HUTTENLOCHER D P. Pictorial structures for object recognition[J]. International Journal of Computer Vision, 2005, 61(1): 55-79.

基于强监督部件模型的遥感图像目标检测

Object detection in remote sensing imagery based on strongly-supervised deformable part models

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

[1]	马佳良, 陈斌, 孙晓飞. 基于改进的Faster R-CNN的通用目标检测框架[J]. 计算机应用, 2021, 41(9): 2712-2719.
[2]	宋中山, 梁家锐, 郑禄, 刘振宇, 帖军. 基于双向门控尺度特征融合的遥感场景分类[J]. 计算机应用, 2021, 41(9): 2726-2735.
[3]	徐江浪, 李林燕, 万新军, 胡伏原. 结合目标检测的室内场景识别方法[J]. 计算机应用, 2021, 41(9): 2720-2725.
[4]	陈静, 毛莺池, 陈豪, 王龙宝, 王子成. 基于改进单点多盒检测器的大坝缺陷目标检测方法[J]. 计算机应用, 2021, 41(8): 2366-2372.
[5]	孙潇, 徐金东. 基于级联生成对抗网络的遥感图像去雾方法[J]. 计算机应用, 2021, 41(8): 2440-2444.
[6]	樊玮, 李晨炫, 邢艳, 黄睿, 彭洪健. 航空发动机损伤图像的二分类到多分类递进式检测网络[J]. 计算机应用, 2021, 41(8): 2352-2357.
[7]	肖振远, 王逸涵, 罗建桥, 熊鹰, 李柏林. 基于部分加权损失函数的RefineDet[J]. 计算机应用, 2021, 41(7): 1928-1932.
[8]	高钦泉, 黄炳城, 刘文哲, 童同. 基于改进CenterNet的竹条表面缺陷检测方法[J]. 计算机应用, 2021, 41(7): 1933-1938.
[9]	侯笑晗, 金国栋, 谭力宁, 薛远亮. 基于自适应和最优特征的合成孔径雷达舰船检测方法[J]. 计算机应用, 2021, 41(7): 2150-2155.
[10]	李朝, 兰海, 魏宪. 基于注意力的毫米波-激光雷达融合目标检测[J]. 计算机应用, 2021, 41(7): 2137-2144.
[11]	吴蕾, 杨晓敏. 改进的基于通道注意力反馈网络的遥感图像融合算法[J]. 计算机应用, 2021, 41(4): 1172-1178.
[12]	后云龙, 朱磊, 陈琴, 吕燧栋. 基于高斯差分特征网络的显著目标检测[J]. 计算机应用, 2021, 41(3): 706-713.
[13]	蒋宁, 方景龙, 杨庆. 基于单点多盒检测器的全局-局部层级的域适应目标检测[J]. 计算机应用, 2021, 41(2): 517-522.
[14]	曹建荣, 吕俊杰, 武欣莹, 张旭, 杨红娟. 融合运动特征和深度学习的跌倒检测算法[J]. 计算机应用, 2021, 41(2): 583-589.
[15]	朱槐雨, 李博. 单阶段多框检测器无人机航拍目标识别方法[J]. 《计算机应用》唯一官方网站, 2021, 41(11): 3234-3241.