Summary

Proceedings of the 2012 International Symposium on Nonlinear Theory and its Applications

2012

Session Number:A2L-C

Session:

Number:110

Robust Human Detector based on HLAC and HOG using RGB-D

Miho Morita,  Hiroshi Takemura,  Hiroshi Mizoguchi,  

pp.110-113

Publication Date:

Online ISSN:2188-5079

DOI:10.15248/proc.1.110

PDF download (675.5KB)

This paper proposes a new human detection method using RGB-D data. RGB-D data include RGB images and Depth (stereo) images. Stereo image can represent human's three-dimensional shape. This method is base integration with two features: Higher-order Local Auto-Correlation (HLAC) features and Histogram of Oriented Gradients (HOG) features. We extend HLAC features from RGB images and HOG features from Stereo images. HLAC features can give a broad pattern of gray scale image. HOG features can give an accurate description of contour of human body. To use Stereo images, HOG features can give a contour more accurate than conventional method. We use co-occurrence of multiple features to integrate HOG and HLAC features, called IHH. In our experiments, we obtain 12.0% lower value on false positive per window than other proposed IHH method when miss rates are similar. These results proved the effectiveness of this new method.

[1] K. Levi et al., “Learning Object Detection from a Small Number of Example: the Importance of Good Features,” n Proc. IEEE Conf. on CVPR, vol. 2, pp. 53-60, 2004.

[2] K. Hotta et al., “Scale Invariant Face Detection Method using Higher-Order Local Auttocorrelation Features extracted from Log-Polar Image”, AFGR, pp. 70-75, 1998.

[3] N. Dalal et al., “Histograms of Oriented Gradients for Human Detection”, CVPR, pp.886-893, 2005.

[4] W. Lin, M. Oakes and J. Tait, “Real AdaBoost for Large Vocabulary Image Classification”, CBMI pp.192-199, 2008.

[5] A. Hidaka et al., “Object Detection by Integration of HLAC Mask Features”, Proc. Of ECSIC Symposium on BLISS 2008.

[6] T. Mita et al., “Discriminative Feature Co-Occurrence Selection for Object Detection”, PAMI, vol.30, No.7, pp.1257-1269, 2008.

[7] M. Morita et al.,"Human Detection Method Based on Feature Co-occurrence of HLAC and HOG" Proceedings of the 2011 IEEE ROBIO, pp.1689-1694, 2011.

[8] M. Enzweiler et al., ”Multi-Cue Pedestrian Classification with Partial Occlusion Handling”, IEEE Conference on CVPR, pp.990-997, 2010.

[9] H. Hirschmueller et al., “Stereo Processing by Semi-global Matching and mutual Information,” IEEE Trans Pattern Anal Mach Intel. , 30(2) pp.328-341, 2008.

[10] A. Martin et al., “The DET Curve in Assessment of Detection Task Performance”, Proc. Eurospeech '97, vol. 4, pp. 1899-1903, 1997.