E-mail Alert
RSS
-
摘要
传统的基于HOG与LBP的特征融合行人检测方法光谱信息损失多、对噪声较为敏感，原始的LBP算法对不均匀的光照变化鲁棒性差，对纹理特征的旋转不变性差。为了克服以上缺点，本文提出了一种基于CLBC和HOG特征融合的行人检测算法。首先，计算原始图像的CLBC特征，并计算基于CLBC纹理特征谱的HOG特征。接着计算原始图像的HOG特征以提取图像的边缘特征。然后将图像的三种特征融合来描述图像，并使用PCA方法降低特征维度，最后使用HIKSVM分类器实现最终对行人的检测。本文分别在Caltech行人数据库和INRIA行人数据库进行实验以验证所提出算法的有效性。实验结果表明，本文所提出的算法有效地提高了行人检测的精度。
Abstract
The traditional feature fusion method based on HOG and LBP loses much spectral information, and it is more sensitive to noise. The original LBP algorithm has poor robustness to uneven illumination changes and poor rotation invariance to texture features. In order to overcome these shortcomings of the method, this paper proposes a pedestrian detection algorithm based on the feature fusion of CLBC and HOG. First, the CLBC feature of the original image is calculated, and the HOG feature based on the CLBC texture feature spectrum is calculated. The HOG feature of the original image is then calculated to extract the edge feature of the image. Then three features of the image are fused to describe the image, and after that we use principal component analysis to reduce the feature dimension. Finally, the detection of the pedestrian is realized by using the HIKSVM classifier. In this paper, experiments are carried out in Caltech pedestrian database and INRIA pedestrian database to verify the effectiveness of the proposed algorithm. The final experimental results show that the proposed algorithm improves the accuracy of pedestrian detection.
-
Key words:
- pedestrian detection /
- HOG /
- CLBC /
- feature extraction /
- feature fusion
-
Overview
Overview: Pedestrian detection is widely used in the field of computer vision, such as public security, intelligent robots, visual surveillance and behavior analysis and so on. However, due to the various factors like complex and changeable environment, different shooting angles, diversity of human behavior, pedestrian detection accuracy and efficiency are not high in the practical application. Therefore, the research of pedestrian detection algorithm is still an important topic in the field of computer vision. Pedestrian detection can generally be considered as the combination of feature extraction with classifier design to automatically detect an existing object from an unknown image or video. With the concept of deep learning proposed, more and more deep learning algorithms have been applied to pedestrian detection. In the pedestrian detection system, the detection mode which combines the HOG features with the LBP features and classifies them with the HIKSVM classifier has been widely used and has achieved good results. HOG features and LBP features have been widely used in feature extraction, at the same time, more and more experts and scholars are also committed to optimizing the existing features. HOG and its improved algorithm obtains good experimental results. However, due to the nature of the gradient, the HOG descriptor is quite sensitive to noise. LBP is a simple but effective operator for describing local image modes. Many of its improved operators are also proposed for extracting the texture features of an image. However, the original LBP and the improved LBP operator are ineffective in extracting the local gray-level difference information, and have the problems of poor robustness to the noise and poor rotation invariance. In order to overcome these shortcomings of the method, this paper proposes a pedestrian detection algorithm based on the feature fusion of CLBC and HOG. First, the CLBC feature of the original image is calculated, and the HOG feature based on the CLBC texture feature spectrum is calculated. The HOG feature of the original image is then calculated to extract the edge feature of the image. Then three features of the image are fused to describe the image, and after that we use principal component analysis to reduce the feature dimension. Finally the detection of the pedestrian is realized by using the HIKSVM classifier. In this paper, experiments are carried out in Caltech pedestrian database and INRIA pedestrian database to verify the effectiveness of the proposed algorithm. The final experimental results show that the proposed algorithm improves the accuracy of pedestrian detection.
-
-
图 1 基于改进HOG-CLBC的行人检测算法流程图
Figure 1. Flow chart of extended HOG-CLBC for pedstrain detection
表 1 Caltech数据集下6种算法的miss rate值
Table 1. The miss rate value of the six algorithms under the Caltech dataset
行人检测算法 VJ HOG MultiFtr HOG-LBP CoHLBP HOG-CLBC Miss rate/% 95 68 68 68 63 61 
下载: 导出CSV
表 2 INRIA数据集下3种算法的分类结果对比
Table 2. Comparison of the classification results of 3 algorithms under the INRIA dataset
行人检测算法 分类率/% 每幅特征提取时间/s HOG 89.08 0.739 HOG-LBP 93.07 0.767 HOG-CLBC 98.58 0.777
下载: 导出CSV
表 3 INRIA数据集下6种算法的miss rate值
Table 3. The miss rate value of the six algorithms under the INRIA dataset
行人检测算法 VJ HOG MultiFtr HOG-LBP CoHLBP HOG-CLBC Miss rate/% 72 46 36 39 25 20
下载: 导出CSV
-
参考文献
[1] 苏松志, 李绍滋, 陈淑媛, 等.行人检测技术综述[J].电子学报, 2012, 40(4): 814-820. http://www.cnki.com.cn/Article/CJFDTOTAL-DZXU201204030.htm
Su S Z, Li S Z, Chen S Y, et al. A survey on pedestrian detection[J]. Acta Electronica Sinica, 2012, 40(4): 814-820. http://www.cnki.com.cn/Article/CJFDTOTAL-DZXU201204030.htm
[2] Benenson R, Omran M, Hosang J, et al. Ten years of pedestrian detection, what have we learned?[C]//Proceedings ofComputer Vision - ECCV 2014 Workshops, 2014: 613-627.
http://www.springerlink.com/openurl.asp?id=doi:10.1007/978-3-319-16181-5_47 [3] Dollar P, Wojek C, Schiele B, et al. Pedestrian detection: an evaluation of the state of the art[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(4): 743-761. doi: 10.1109/TPAMI.2011.155
[4] Ouyang W L, Wang X G. Joint deep learning for pedestrian detection[C]//Proceedings of IEEE International Conference on Computer Vision, 2013: 2056-2063.
http://ieeexplore.ieee.org/document/6751366 [5] Zeng X Y, Ouyang W L, Wang X G. Multi-stage contextual deep learning for pedestrian detection[C]//Proceedings of IEEE International Conference on Computer Vision, 2013: 121-128.
http://ieeexplore.ieee.org/xpls/icp.jsp?arnumber=6751124 [6] Luo P, Tian Y L, Wang X G, et al. Switchable deep network for pedestrian detection[C]//Proceedings of Conference on Computer Vision and Pattern Recognition, 2014: 899-906.
http://doi.ieeecomputersociety.org/10.1109/CVPR.2014.120 [7] Zeng X Y, Ouyang W L, Wang M, et al. Deep learning of scene-specific classifier for pedestrian detection[C]//Fleet D, Pajdla T, Schiele B, et al. Computer Vision - ECCV 2014. Cham: Springer, 2014: 472-487.
http://link.springer.com/10.1007/978-3-319-10578-9_31 [8] Dalal N, Triggs B. Histograms of oriented gradients for human detection[C]//Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2005: 886-893.
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=1467360 [9] Ojala T, Pietikäinen M, Harwood I. A comparative study of texture measures with classification based on featured distributions[J]. Pattern Recognition, 1996, 29(1): 51-59. doi: 10.1016/0031-3203(95)00067-4
[10] Tribak H, Moughyt S, Zaz Y, et al. Remote QR code recognition based on HOG and SVM classifiers[C]//Proceedings of International Conference on Informatics and Computing, 2017: 137-141.
http://ieeexplore.ieee.org/document/7905704/ [11] Wang X Y, Han T X, Yan S C. An HOG-LBP human detector with partial occlusion handling[C]//Proceedings of IEEE 12th International Conference on Computer Vision, 2010: 32-39.
http://ci.nii.ac.jp/naid/10029956697 [12] Fan G J, Li B, Mu W Q, et al. HOGG: Gabor and hog-based human detection[C]//Proceedings of 8th International Conference on Information Technology in Medicine and Education, 2016: 562-566.
http://en.cnki.com.cn/Article_en/CJFDTotal-WXJY201621005.htm [13] 彭志蓉, 赵美蓉, 杨伟明, 等.改进的多光谱聚合通道行人检测[J].光电工程, 2017, 44(9): 882-887. doi: 10.3969/j.issn.1003-501X.2017.09.004
Peng Z R, Zhao M R, Yang W M, et al. Improved multispectral aggregate channel for pedestrian detection[J]. Opto-Electronic Engineering, 2017, 44(9): 882-887. doi: 10.3969/j.issn.1003-501X.2017.09.004
[14] Zhao L H, Liu F, Wang Y J. Face recognition based on LBP and genetic algorithm[C]//Proceedings of Chinese Control and Decision Conference, 2016: 1582-1587.
http://ieeexplore.ieee.org/document/7531236/ [15] Liu Y C, Huang S S, Lu C H, et al. Thermal pedestrian detection using block LBP with multi-level classifier[C]//Proceedings of International Conference on Applied System Innovation, 2017: 602-605.
http://ieeexplore.ieee.org/document/7988495/ [16] Guo Z H, Zhang L, Zhang D. A completed modeling of local binary pattern operator for texture classification[J]. IEEE Transactions on Image Processing, 2010, 19(6): 1657-1663. doi: 10.1109/TIP.2010.2044957
[17] 程卫亮, 王向军, 万子敬, 等.压缩域目标跟踪算法在小型化DSP平台上的研究与实现[J].光电工程, 2017, 44(10): 972-982. doi: 10.3969/j.issn.1003-501X.2017.10.005
Cheng W L, Wang X J, Wan Z J, et al. Research and implementation of target tracking algorithm in compression domain on miniaturized DSP platform[J]. Opto-Electronic Engineering, 2017, 44(10): 972-982. doi: 10.3969/j.issn.1003-501X.2017.10.005
[18] Ji L P, Ren Y, Liu G S, et al. Training-based gradient lbp feature models for multiresolution texture classification[J]. IEEE Transactions on Cybernetics, 2017, pp(99): 1-14. http://cn.bing.com/academic/profile?id=449e89975875dc793038d766a42a7722&encoded=0&v=paper_preview&mkt=zh-cn
[19] Zhao Y, Huang D S, Jia W. Completed local binary count for rotation invariant texture classification[J]. IEEE Transactions on Image Processing, 2012, 21(10): 4492-4497. doi: 10.1109/TIP.2012.2204271
[20] Maji S, Berg A C, Malik J. Classification using intersection kernel support vector machines is efficient[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2008: 1-8.
http://www.mendeley.com/research/classification-using-intersection-kernel-support-vectormachines-efficien/ [21] Wojek C, Schiele B. A performance evaluation of single and multi-feature people detection[C]//Proceedings of the 30th DAGM symposium on Pattern Recognition, 2008: 82-91.
http://link.springer.com/chapter/10.1007/978-3-540-69321-5_9 [22] Watanabe T, Ito S. Two co-occurrence histogram features using gradient orientations and local binary patterns for pedestrian detection[C]//Proceedings of 2nd IAPR Asian Conference on Pattern Recognition, 2013: 415-419.
[23] Viola P, Jones M J. Robust real-time face detection[J]. International Journal of Computer Vision, 2004, 57(2): 137-154. doi: 10.1023/B:VISI.0000013087.49260.fb
-
访问统计
点击扫一扫
图(9)
表(3)
计量
- 文章访问数:
- PDF下载数:
- 施引文献: 0
