Journal of the Korea Convergence Society (한국융합학회논문지)

Korea Convergence Society (한국융합학회)
권호 표지
DOI QR코드

DOI QR Code

Modified HOG Feature Extraction for Pedestrian Tracking

동영상에서 보행자 추적을 위한 변형된 HOG 특징 추출에 관한 연구

  • Kim, Hoi-Jun (Dept of Plasma Bio Display, KwangWoon University) ;
  • Park, Young-Soo (Ingenium College of Liberal Arts, KwangWoon University) ;
  • Kim, Ki-Bong (Department of computer information, Daejeon health institute of technology) ;
  • Lee, Sang-Hun (Ingenium College of Liberal Arts, KwangWoon University)
  • 김회준 (광운대학교 플라즈마바이오디스플레이학과) ;
  • 박영수 (광운대학교 인제니움학부대학) ;
  • 김기봉 (대전보건대학 컴퓨터정보학과) ;
  • 이상훈 (광운대학교 인제니움학부대학)
  • Received : 2019.01.11
  • Accepted : 2019.03.20
  • Published : 2019.03.28
Download PDF
⟨ Previous Next ⟩

Abstract

In this paper, we proposed extracting modified Histogram of Oriented Gradients (HOG) features using background removal when tracking pedestrians in real time. HOG feature extraction has a problem of slow processing speed due to large computation amount. Background removal has been studied to improve computation reductions and tracking rate. Area removal was carried out using S and V channels in HSV color space to reduce feature extraction in unnecessary areas. The average S and V channels of the video were removed and the input video was totally dark, so that the object tracking may fail. Histogram equalization was performed to prevent this case. HOG features extracted from the removed region are reduced, and processing speed and tracking rates were improved by extracting clear HOG features. In this experiment, we experimented with videos with a large number of pedestrians or one pedestrian, complicated videos with backgrounds, and videos with severe tremors. Compared with the existing HOG-SVM method, the proposed method improved the processing speed by 41.84% and the error rate was reduced by 52.29%.

본 논문에서는 실시간으로 보행자를 추적할 때 배경 제거를 이용하여 변형된 HOG(Histogram of Oriented Gradients) 특징 추출을 제안하였다. 기존의 HOG 특징 추출은 연산량이 많아 추적 속도가 느린 문제가 있다. 배경 제거를 통해 연산량 감소와 추적률을 향상시키기 위해 연구하였다. 불필요한 영역에서의 특징 추출을 감소시키기 위해 HSV 색공간에서 S 채널과 V 채널을 이용하여 영역 제거를 진행하였다. 영상의 평균 S 채널과 V 채널로 배경 제거 후 입력 영상이 전체적으로 어두워 객체 추적에 실패하는 경우가 있다. 이러한 경우를 방지하기 위해 히스토그램 평활화를 하였다. 제거된 영역에서 추출되는 HOG 특징이 감소되고, 객체에서는 명확한 HOG 특징이 추출되어 객체 추적 속도와 추적률이 향상되었다. 본 실험에서는 다수의 보행자나 한명의 보행자가 존재하는 영상, 배경이 복잡한 영상, 흔들림이 심한 영상을 가지고 실험하였다. 제안하는 방법은 기존의 HOG-SVM 방법과 비교하여 처리 속도는 약 41.84% 향상되었으며 오 추적률은 약 52.29% 감소되어 개선된 추적률을 보였다.

Keywords

OHHGBW_2019_v10n3_39_f0001.png 이미지

Fig. 1. HSV Color Coordinate (a) HSV Cylinder (b) HSV Cone

OHHGBW_2019_v10n3_39_f0002.png 이미지

Fig. 2. Image of gradient direction (a) Original Image (b) Gradient direction

OHHGBW_2019_v10n3_39_f0003.png 이미지

Fig. 3. Margin and Support Vector

OHHGBW_2019_v10n3_39_f0004.png 이미지

Fig. 4. Flowchart of the Proposed Method

OHHGBW_2019_v10n3_39_f0005.png 이미지

Fig. 5. Remove Background(HSV) (a) Original Video (b) Remove background

OHHGBW_2019_v10n3_39_f0006.png 이미지

Fig. 6. Difficult to Distinguish between Objects (a) Remove Background Video (b) Indistinguishable Objects

OHHGBW_2019_v10n3_39_f0007.png 이미지

Fig. 7. Remove Background and Equalization (a) Gray Video and Histogram (b) Equalization Video and Histogram

OHHGBW_2019_v10n3_39_f0008.png 이미지

Fig. 8. Distinct objects of equalization (a) Equalization Video (b) Distinct Objects

OHHGBW_2019_v10n3_39_f0009.png 이미지

Fig. 9. HOG Feature Video (a) HOG Feature of Original Video (b) HOG Feature of Proposed Method

OHHGBW_2019_v10n3_39_f0010.png 이미지

Fig. 10. HOG-SVM Tracking Result (a) HOG Feature Video (b) SVM Tracking Video

OHHGBW_2019_v10n3_39_f0011.png 이미지

Fig. 11. Experiment Results 1 (Walking1) (a) Original Video (b) HOG-SVM Tracking (c) Proposed Method Tracking

OHHGBW_2019_v10n3_39_f0012.png 이미지

Fig. 12. Experiment Results 2 (Jogging2) (a) Original Video(b) HOG-SVM Tracking (c) Proposed Method Tracking

OHHGBW_2019_v10n3_39_f0013.png 이미지

Fig. 13. Experiment Results 3 (Walking2) (a) Original Video(b) HOG-SVM Tracking (c) Proposed Method Tracking

OHHGBW_2019_v10n3_39_f0014.png 이미지

Fig. 14. Experiment Results 4 (Human8, Couple) (a) Original Video (b) HOG-SVM Tracking (c) Proposed Method Tracking

OHHGBW_2019_v10n3_39_f0015.png 이미지

Fig. 15. Experiment Results 5 (David3, Human2, 7) (a) Original Video (b) HOG-SVM Tracking (c) Proposed Method Tracking

OHHGBW_2019_v10n3_39_f0016.png 이미지

Fig. 16. Experiment Results 6 (Human9) (a) Original Video (b) HOG-SVM Tracking (c) Proposed Method Tracking

Table 1. Tracking Processing Speed (FPS, ms)

OHHGBW_2019_v10n3_39_t0001.png 이미지

Table 2. Average for Consecutive Error Tracking Frame

OHHGBW_2019_v10n3_39_t0002.png 이미지

References

  1. H. Song, Y. Zheng & K. Zhang. (2016). Robust visual tracking via self-similarity learning. Electronics Letters, 53(1), 20-22. DOI : 10.1049/el.2016.3011
  2. J. Y. Kim. (2015). Method on Detection Specific Region Using License Plate Edge Feature In Car Image. Master dissertation. Kwangwoon University, Seoul.
  3. J. H. Park. (2017). Improved face detection algorithm using color distribution and shape characteristics. Master dissertation. Kwangwoon University, seoul.
  4. S. K. Pyo, G. S. Lee, Y. S. Park, S. H. Lee. (2018). A license plate detection method based on contour extraction that adapts to environmental changes. Journal of the Korea Convergence Society, 9(9), 31-39. https://doi.org/10.15207/JKCS.2018.9.9.031
  5. M. Cen & C. Jung. (2017). Complex Form of Local Orientation Plane for Visual Object Tracking. IEEE Access, 5, 21597-21604. DOI : 10.1109/ACCESS.2017.2756699
  6. L. Lan, X. Wang, S. Zhang, D. Tao, W. Gao & T. S. Huang. (2018). Interacting Tracklets for Multi-Object Tracking. IEEE Transactions on Image Processing. 27(9). 4585-4597 DOI : 10.1109/TIP.2018.2843129
  7. R. Yu, I.. Cheng, B. Zhu, S. Bedmutha & A. Basu. (2018). Adaptive Resolution Optimization and Tracklet Reliability Assessment for Efficient Multi-Object Tracking. IEEE Transactions on Circuits and Systems for Video Technology. 28(7). 1623-1633. DOI : 10.1109/TCSVT.2017.2668278
  8. Q. Xie, O. Remil, Y. Guo, M. Wang, M. Wei & J. Wang. (2018). Object Detection and Tracking Under Occlusion for Object-Level RGB-D Video Segmentation. IEEE Transactions on Multimedia. 20(3). 580-592. DOI : 10.1109/TMM.2017.2751965
  9. T. H. Yoo, G. S. Lee & S. H. Lee. (2012). Window Production Method based on Low-Frequency Detection for Automatic Object Extraction of GrabCut. Journal of Digital Convergence, 10(8), 211-217 https://doi.org/10.14400/JDPM.2012.10.8.211
  10. H. H. Han, G. S. Lee, J. Y. Lee & S. H. Lee. (2012). Region Segmentation Technique Based on Active Contour for Object Segmentation, Journal of Digital Convergence, 10(3), 167-172. https://doi.org/10.14400/JDPM.2012.10.3.167
  11. J. C. Han, B. C. Koo & K. J. Cheoi. (2017). Obstacle Detection and Recognition System for Autonomous Driving Vehicle. Journal of Convergence for Information Technology, 7(6), 229-235. DOI : 10.22156/CS4SMB.2017.7.6.229
  12. M. K. Kwon & H. S. Yang. (2017). A scene search method based on principal character identification using convolutional neural network. Journal of Convergence for Information Technology, 7(2), 31-36. DOI : 10.22156/CS4SMB.2017.7.2.031
  13. A. H. Ahmed, K. Kpalma & A. O. Guedi. (2017, December). Human Detection Using HOG-SVM, Mixture of Gaussian and Background Contours Subtraction. In 2017 13th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS) (pp. 334-338). IEEE.
  14. B. Meus, T. Kryjak & M. Gorgon. (2017, September). Embedded vision system for pedestrian detection based on HOG+SVM and use of motion information implemented in Zynq heterogeneous device. In Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA), (pp. 406-411). IEEE.
  15. J. H. Kim. (2015). A Study on Tracking in Video Using Modified Particle Filter. Master dissertation. Kwangwoon University, Seoul.
  16. N. Dalal & B. Triggs. (2005, June). Histograms of oriented gradients for human detection. In Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society Conference on (Vol. 1, pp. 886-893). IEEE.
  17. C. Cortes, & V. Vapnik. (1995). Support-vector networks. Machine learning, 20(3), 273-297. https://doi.org/10.1007/BF00994018
  18. J. H. Baek, J. S. Kim & E. T. Kim. (2014, May). Pedestrian detection with NIR sensor-based HOG-SVM. Institude of Control, Robotics and Systems. (pp. 650-651).
  19. Daimler Mono Ped. Detection Benchmark Data set. http://www.gavrila.net