收稿日期: 2017-01-02
修回日期: 2017-03-05
网络出版日期: 2018-05-31
基金资助
国家自然科学基金（No.41101410）资助
Target Detection from High Resolution Remote Sensing Images Based on Combination of Multi-scale Visual Features
Received date: 2017-01-02
Revised date: 2017-03-05
Online published: 2018-05-31
针对高空间分辨率（高分）遥感影像中目标空间结构明显、颜色丰富的特点，提出一种基于多尺度视觉特征组合的目标检测方法.首先采用两组特征描述子来表达目标的颜色与空间结构特征，其中，一组是彩色变换后的密集空间金字塔尺度不变特征变换（scale-invariantfeature transform，SIFT）描述子，另一组是多尺度方向梯度直方图（histogram of orientedgradient，HOG）与色度、饱和度、明度（hue，saturation and value，HSV）颜色模型组合的特征描述子，然后用支持向量机（support vector machine，SVM）对两组特征分别进行模型训练与目标检测，对于彩色变换密集空间金字塔SIFT描述子，采用视觉词袋空间金字塔匹配核（spatial pyramid matching kernel，SPMK）构建SVM的输入，对于HOG与HSV特征，直接将特征向量按顺序合成作为SVM的输入，最后取两组检测结果的交集作为最终检测结果.为验证方法的有效性，分别以航空影像中的建筑物和鱼排，以及高分卫星影像中的游艇为目标进行检测.结果表明，该方法的目标检测精度可达到90%以上.
冯发杰, 吏军平, 丁亚洲, 朱坤, 崔卫红 . 基于多尺度视觉特征组合的高分遥感影像目标检测[J]. 应用科学学报, 2018 , 36(3) : 471 -484 . DOI: 10.3969/j.issn.0255-8297.2018.03.007
To utilize the spatial structure and color information of the target in the high resolution remote sensing images, this paper proposes a method of target detection based on multi-scale visual feature combination. Firstly, we use two types of feature descriptor to express the color and spatial structure information of targets, one is the dense spatial pyramid scale-invariant feature transform (SIFT) descriptor in spatial pyramid of the color transformed images, the other is the multi-scale histogram of oriented gradient (HOG) feature combined with the hue, saturability and value (HSV) features. Secondly, the two sets of features are used to train classifer through support vector machine (SVM). For the color transformed dense SIFT spatial pyramid feature descriptor, the input of SVM is constructed by spatial pyramid matching kernel (SPMK). For the combined HOG and HSV feature, the orderly-combined eigenvectors are taken as the input of the SVM classifer. Finally the intersection of the two classifcation results are regarded as the fnal results. For verifcation, we tested the buildings and fsh rafts in aerial images and the yachts in satellite images. Experimental results show that the proposed method can improve the precision of target detection, which can reach to by 90% or better.
[1] Blaschke T. Object based image analysis for remote sensing[J]. ISPRS Journal of Photogrammetry & Remote Sensing, 2010, 65(1):2-16.
[2] 周军其,李志娟. 空间关系辅助的面向对象建筑物提取[J]. 应用科学学报,2012, 30(5):511-526. Zhou J Q, Li Z J. Spatial relation-aided method for object-oriented extraction of buildings from high resolution image[J]. Journal of Applied Sciences, 2012, 30(5):511-526. (in Chinese)
[3] 何少林,徐京华,张帅毅. 面向对象的多尺度无人机影像土地利用信息提取[J]. 国土资源遥感,2013, 25(2):107-112. He S L, Xu J H, Zhang S Y. Land use classifcation of object-oriented multi-scale by UAV image[J]. Remote Sensing for Land and Resources, 2013, 25(2):107-112. (in Chinese)
[4] 任慧,谈亚光. 基于面向对象的国情普查地表覆盖提取方法[J]. 地理空间信息,2014, 12(3):14-16. Ren H, Tan Y G. Object-oriented surface coverage extract methods of geographic situation census[J]. Geospatial Information, 2014, 12(3):14-16. (in Chinese)
[5] 汤育红. 地理国情普查地表覆盖与国情要素信息的提取方法探讨[J]. 测绘与空间地理信息,2013, 36(12):89-91. Tang Y H. Geographical conditions census surface coverage with the state of the element information extraction method to explore[J]. Geomatics & Spatial Information Technology, 2013, 36(12):89-91. (in Chinese)
[6] 翟亮,张晓贺,桑会勇. 面向地理国情普查的地表覆盖分类技术与试验[J]. 遥感信息,2014(4):71-75. Zhai L, Zhang X H, Sang H Y. Land cover classifcation for national geographic conditions gensu[J]. Remote Sensing Information, 2014(4):71-75. (in Chinese)
[7] 黄慧萍. 面向对象影像分析中的尺度问题研究[D]. 北京:中国科学院遥感应用研究所,2003.
[8] 林先成,李永树. 成都平原高分辨率遥感影像分割尺度研究[J]. 国土资源遥感,2010, 2010(2):7-11. Lin X C, Li Y S. A study of the segmentation scale of high-resolution remotely sensed data in Chengdu plain[J]. Remote Sensing for Land & Resources, 2010(2):7-11. (in Chinese)
[9] Meng Y, Lin C, Cui W. Scale selection based on Moran's I for segmentation of high resolution remotely sensed images[C]//IGARSS IEEE International Geoscience & Remote Sensing Symposium. IEEE, 2014:4895-4898.
[10] Sivic J, Zisserman A. Video Google:a text retrieval approach to object matching in videos[C]//Proceedings of the 9th IEEE International Conference on Computer Vision, Nice, France, 2003:1470-1477.
[11] Lazebnik S, Schmid C, Ponce J. Beyond bags of feature:spatial pyramid matching for recognizing natural scene categories[C]//Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR). New York:IEEE, 2006:2169-2178.
[12] Yang Y, Newsam S. Spatial pyramid co-occurrence for image classifcation[J]. IEEE International Conference on Computer Vision, 2011, 23(5):1465-1472.
[13] Mazinan A H, Sarikhani M. Providing an efcient intelligent transportation system through detection, tracking and recognition of the region of interest in trafc signs by using non-linear SVM classifer in line with histogram oriented gradient and Kalman flter approach[J]. Sadhana. 2014, 39(1):27-37.
[14] Moranduzzo T, Melgani F. Detecting cars in UAV images with a catalog-based approach[J]. IEEE Transactions on Geoscience and Remote Sensing, 2014, 52(10):6356-6367.
[15] Lowe D G. Object recognition from local scale-invariant features[C]//The proceedings of the seventh IEEE international conference on Computer vision, 1999(2):1150-1157.
[16] Lowe D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004, 60(2):91-110.
[17] Dalal N, Triggs B. Histograms of oriented gradients for human detection[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2005:886-893.
[18] Sande K E A, Gevers T, Snoek C G M. Evaluating color descriptors for object and scene recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2010, 32(9):1582-1596.
[19] Corinna C, Vladimir V. Support-vector networks[J]. Machine Learning, 1995, 20(3):273- 297.
[20] Grauman, K. Darrell, T. The pyramid match kernel:discriminative classifcation with sets of image features[C]//10th IEEE International Conference on Computer Vision (2005 ICCV), 2005(2):1458-1465.
/
| 〈 |
|
〉 |
