Çoklu Yaya Takip Probleminin Grafik İşleme Biriminde Gerçekleştirilmesi

Year 2021, Volume: 01 Issue: 01, 26 - 33, 17.07.2021

Özcan Dülger

Abstract

Parçacık süzgeci uygulamanın sistem yada ölçüm modellerinin oldukça doğrusal olmadığı ve gürültülerin büyük olduğu durumlarda kullanılan bir seri Monte Carlo kestirim yöntemidir. Parçacık sayısı arttıkça parçacık süzgecinin hesaplama maliyetleri artmaktadır. Grafik işleme birimi içerdiği çok sayıdaki çekirdek ile parçacık süzgecini hızlandırmak için ümit verici çözümler sunmaktadır. Çoklu yaya takip probleminde çok sayıda yaya olduğu için birden çok parçacık süzgeci aynı anda çalışmaktadır. Bu yüzden parçacık süzgecini grafik işleme biriminde verimli şekilde çalıştırmak önemli olmaktadır. Bu çalışmada, çoklu yaya takip yöntemini grafik işleme biriminde gerçekleştiriyoruz. 10 tane yanlış alarm (yayaya ait olmayan ölçümler) ve 3 tane yaya olmak üzere bir zaman adımında en çok 13 tane ölçüm değeri alıcıdan gelmektedir ve en az 10 ve en çok 13 tane parçacık süzgeci aynı anda çalışmaktadır. Bu parçacık süzgeçlerini grafik işleme biriminde çalıştırdık ve 9.79x’e kadar hızlanma elde ettik. İki ardışık ölçümlerin arasındaki sürenin çok kısa olduğunu düşünürsek parçacık sayısı arttıkça parçacık süzgecinin grafik işleme biriminde çalıştırılmasının önemini daha iyi görebilmekteyiz. Ayrıca deneylerde elde ettiğimiz kalite sonuçları da kayda değer çıkmıştır.

Keywords

çoklu yaya takibi, parçacık süzgeci, Sistematik yeniden öğrenme, grafik işleme birimi, CUDA programlama, Tesla K40 ekran kartı

Implementation of the Multi Pedestrian Tracking Problem on Graphics Processing Unit

Abstract

Particle filter is a serial Monte Carlo estimation method which is used when the system or the measurement model of the application is highly non-linear and uncertainties are large. As the number of particles increases, the computation cost of the particle filter increases. Graphics processing unit (GPU) offers promising solutions to accelerate the particle filter. Since there are many pedestrians in a multi pedestrian tracking problem, more than one particle filters run at a time. So, it is important to implement the particle filter on the GPU efficiently. In this study, we implement a multi pedestrian tracking algorithm on the GPU. We have three pedestrians along with some clutters (ten clutters at each time step). There may be up to 13 measurements which stand for too many particle filters run at a time. We use gating, association techniques in order to assign a measurement to a track (a pedestrian). We implement the particle filters on the GPU and achieve up to 9.79x speed up. When we consider the duration between two consecutive measurements is small, implementing the particle filter on the GPU becomes substantial as the number of particles increases. Furthermore, the quality of the particle filters is significant.

Keywords

multi pedestrian tracking, particle filter, Systematic resampling, graphics processing unit, CUDA programming, Tesla K40 board

References

• [1] B. Ristic, S. Arulampalam, and N. Gordon, Beyond the Kalman Filter Particle Filters for Tracking Applications. Artech House, 2004.
• [2] P. Gong, J. D. Basciftci, and F. Ozguner, "A Parallel Resampling Algorithm for Particle Filtering on Shared-Memory Architectures," Parallel and Distributed Processing Symposium Workshops & PhD Forum (IPDPSW), 2012 IEEE 26th International, Shanghai, pp. 1477-1483, 2012.
• [3] D. Brscic, T. Kanda, T. Ikeda, and T. Miyashita, "Person position and body direction tracking in large public spaces using 3D range sensors," IEEE Transactions on Human-Machine Systems, Vol. 43, No. 6, pp. 522-534, 2013.
• [4] G. Hendeby, J. D. Hol, R. Karlsson, and F. Gustafsson, "A graphics processing unit implementation of the particle filter," In Signal Processing Conference, 2007 15th European, pp. 1639–1643, IEEE, 2007.
• [5] K. Hwang, and W. Sung, "Load balanced resampling for real-time particle filtering on graphics processing units," IEEE Transactions on Signal Processing, Vol. 61, No. 2, pp. 411–419, 2013.
• [6] Y. Wu, J. Wang, and Y. Cao, "Particle filter based on iterated importance density function and parallel resampling," Journal of Central South University, Vol. 22, No. 9, pp. 3427–3439, 2015.
• [7] M. Owczarek, P. Barański, and P. Strumiłło, "Pedestrian tracking in video sequences: a particle filtering approach," In Computer Science and Information Systems (FedCSIS), 2015 Federated Conference on, pp. 875-881, 2015.
• [8] Y. Guan, X. Chen, Y. Wu, and D. Yang, "An improved particle filter approach for real-time pedestrian tracking in surveillance video," In International Conference on Information Science and Technology Applications, Atlantis Press, 2013.
• [9] D. Stadler, and J. Beyerer, "Improving Multiple Pedestrian Tracking by Track Management and Occlusion Handling," Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021.
• [10] F. Flodin, Improved Data Association for Multi-Pedestrian Tracking Using Image Information, Master of Science Thesis in Electrical Engineering Department of Electrical Engineering, Linköping University, 2020.
• [11] L. Barba-Guaman, J. Eugenio Naranjo, and A. Ortiz, "Deep learning framework for vehicle and pedestrian detection in rural roads on an embedded gpu," Electronics, Vol. 9, No. 4, pp. 589, 2020.
• [12] M. Dimitrievski, P. Veelaert, and W. Philips, "Behavioral pedestrian tracking using a camera and lidar sensors on a moving vehicle," Sensors, Vol. 19, No. 2, pp. 391, 2019.
• [13] L. M. Murray, A. Lee, and P. E. Jacob, "Parallel resampling in the particle filter," Journal of Computational and Graphical Statistics, Vol. 25, No. 3, pp. 789–805, 2016.
• [14] S. Blackman, and R. Popoli, Design and Analysis of Modern Tracking Systems. Norwood, MA: Artech House, 1999.
• [15] M. Harris, “Optimizing Parallel Reduction in CUDA, NVIDIA Developer Technology,” 2007. [Online]. Available: http://developer.download.nvidia.com/compute/cuda/1.1-Beta/x86_website/projects/reduction/doc/reduction.pdf
• [16] W. M. Hwu, “A work-eficient parallel scan kernel,” 2014. [Online]. Available: http://ece408.hwu-server2.crhc.illinois.edu/Shared%20Documents/Slides/Lecture-4-6-work-efficient-scan-kernel.pdf
• [17] Intel, “Intel® Core™ i7-4790K Processor,” 2020. [Online]. Available: https://ark.intel.com/content/www/us/en/ark/products/80807/intel-core-i7-4790k-processor-8m-cache-up-to-4-40-ghz.html
• [18] NVIDIA, “Tesla K40 GPU Active Accelerator: Board Specification,” 2013. [Online]. Available: https://www.nvidia.com/content/PDF/kepler/Tesla-K40-Active-Board-Spec-BD-06949-001_v03.pdf
• [19] NVIDIA, “NVIDIA’s Next Generation CUDA Compute Architecture: Kepler GK110/210,” 2014. [Online]. Available: https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/tesla-product-literature/NVIDIA-Kepler-GK110-GK210-Architecture-Whitepaper.pdf