ردیابی تصویری سریع، مطمئن و مقاوم نسبت به انسداد با کمک یک مدل تقسیمشده مبتنی بر لبه
محورهای موضوعی : مهندسی برق و کامپیوترپیمان معلم 1 * , رسول عسگریان دهکردی 2
1 - دانشگاه اصفهان
2 - دانشگاه اصفهان
کلید واژه: رديابی تصويری لبه انسداد بلادرنگ,
چکیده مقاله :
در اين مقاله الگوريتمی سريع، مطمئن و مقاوم نسبت به انسداد براي رديابي تصويری هدفی از پيش مشخص شده در تصاوير متوالي، بر مبناي تطابق قالب لبههاي هدف با لبههاي فضای جستجو ارائه ميگردد. در ابتدا محدوده هدف توسط کاربر مشخص شده و سپس الگوریتم پیشنهادی با انتخاب قویترین لبههای آن، مدلی مناسب برای هدف را مشخص میکند. در ادامه برای افزایش مقاومت نسبت به انسداد، مدل هدف به 4 قسمت تقسیم شده و با AND شدن قالب لبههای هر قسمت با لبههای فضای جستجو و شمارش پیکسلهای غیر صفر آن، ماتریس تشابه برای هر قسمت از هدف به دست میآید. در صورت کمتربودن مقادیر ماتریس تشابه از آستانهای، قسمت مورد نظر مسدود در نظر گرفته شده و در ادامه با در نظر گرفتن تأثیر قسمتهای نامسدود، مکان هدف در هر قاب مشخص میشود. در طی رديابی، در صورت وجود شرایط مناسب با توجه به شرايط پسزمينه، مدل لبههای هدف به روز میگردد. انتخاب قویترین لبهها، چند قسمت کردن و به روز رسانی قالب هدف، مقاومت الگوریتم را نسبت به چالشهایی مانند تغییرات نوری محیط و بروز انسداد بر روی هدف، به همراه امکان تعقيب هدف با دقت بالا را در پی داشته است. سادگی الگوریتم پيشنهادی، امکان پيادهسازی بلادرنگ آن را به زبان C و در محیط OpenCV فراهم کرده است به گونهای که میانگین سرعت آن توسط رایانهای با فرکانس پردازنده GHZ 6/2 و GB RAM4، به بیش از 60 قاب در ثانیه میرسد. مقایسه نتایج این الگوریتم با الگوریتمهای دیگر نشانگر سرعت بسیار بالاتر و قابلیت اطمینان بیشتر الگوریتم پیشنهادی است.
In this paper a fast, reliable and robust algorithm against occlusion for visual tracking of a pre-defined target in sequence images based on adapting template of target edges with search space edges is presented. At first, target window is specified by user and then the proposed algorithm determines an appropriate model for the target by choosing the best edges of the target window. Moreover, to increase robustness against occlusion, target model has been divided into four equal divisions and by performing a logical AND between the template of each division edges and search space edges and then by counting its non-zero pixels, resemblance matrix for each division of target is obtained. In a case that values of the resemblance matrix are less than values of threshold matrix, the desired division is considered occluded and then by taking the effects of non-occluded divisions into account, the exact location of the target in each frame is determined. In the tracking values, in case of appropriate condition respect to background condition, the model of target edges is updated. Selecting dominant edges, multi dividing and updating the target template, has resulted in increasing the robustness of the algorithm against some vital challenges such as changing in ambient and target light, and occurring occlusion over target. The simplicity of this algorithm has provided the possibility of real-time implementation in OpenCV environment using C language, that achieves averagely to more than 60 frames per second for a computer with 2.6 GHz CPU and 4 GB RAM. Moreover, comparing the results of the proposed algorithm to other algorithms, revealed a higher speed and greater reliability.
[1] P. Moallem, A. Memarmoghaddam, and M. Ashourian, "Robust and fast tracking algorithm in video sequences by adaptive window sizing using a novel analysis on spatiotemporal gradient powers," International J. of Circuit, Systems, and Computers, vol. 16, no. 2, pp. 305-317, Apr. 2007.
[2] W. Zhong, H. Lu, and M. Yang, "Robust object tracking via sparsity-based collaborative model," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, CVPR'12, pp. 1838-1845, Jun. 2012.
[3] A. C. Emilio Maggio, Video Tracking: Theory and Practice, Wiley, 2011.
[4] B. Javidi, Image Recognition and Classification, Algorithms, Systems and Applications, 1st Edition, CRC Press, 2002.
[5] T. Zhang, B. Ghanem, S. Liu, and N. Ahuja, "Robust visual tracking via multi-task sparse learning," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, CVPR'12, pp. 2042-2049, Jun. 2012.
[6] L. Cehovin, M. Kristan, and A. Leonardis, "Robust visual tracking using an adaptive coupled-layer visual model," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 35, no. 4, pp. 941-953, Apr. 2013.
[7] X. Jia, D. Wang, and H. Lu, "Fragment-based tracking using online multiple kernel learning," in Proc. 19th IEEE Int. Conf. on Image Processing, ICIP'12, pp. 393-396, Oct. 2012.
[8] J. Wu, H. Yue, Y. Cao, and Z. Cui, "Video object tracking method based on normalized cross-correlation matching," in Proc. 9th Int. Symp. on Distributed Computing and Applications to Business Engineering and Science, DCABES'10, pp. 523-527, Aug. 2010.
[9] C. Chu, J. Hwang, H. Pai, and K. Lan, "Robust video object tracking based on multiple kernels with projected gradients," in Proc. IEEE Int. Conf. on Acoustics, Speech, and Signal Processing, ICASSP'11, pp. 1421-1424, May 2011.
[10] H. Tahvilian, P. Moallem, and A. Monadjemi, "Balloon energy based on parametric active contour and directional walsh-hadamard transform and its application in tracking of texture object in texture background," EURASIP J. on Advances in Signal Processing, Article 253, 15 pp., Dec. 2012.
[11] ر. عسگریان دهکردی، پ. معلم و ع. معمار مقدم، "الگوریتمی قطعه مبنا برای ردیابی بلادرنگ تصویری با مقاومت مناسب نسبت به انسداد و تغییرات هدف،" مجموعه مقالات دومین کنفرانس بینالمللی بازشناسی الگو و تحلیل تصویر، 7 صص.، رشت، اسفند 93.
[12] C. Choi and H. I. Christensen, "Robust 3D visual tracking using particle filtering on the SE (3) group," in Proc. IEEE Int. Conf. on Robotics and Automation, ICRA'11, pp. 4384-4390, May 2011.
[13] C. Choi and H. I. Christensen, "3D texture-less object detection and tracking: an edge-based approach," in Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, IROS'12, pp. 3877-3884, Oct. 2012.
[14] K. Lebeda, S. Hadfield, J. Matas, and R. Bowden, "Long-term tracking through failure cases," in Proc. IEEE Int. Conf. on Computer Vision Workshops, ICCVW'13, pp. 153-160, Dec. 2013.
[15] A. Adam, E. Rivlin, and I. Shimshoni, "Robust fragments-based tracking using the integral histogram," in Proc. IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, vol. 1, pp. 798-805, Jun. 2006.
[16] D. A. Ross, J. Lim, R. Lin, and M. Yang, "Incremental learning for robust visual tracking," International J. of Computer Vision, vol. 77, no. 3, pp. 125-141, May 2008.
[17] D. Comaniciu, V. Ramesh, and P. Meer, "Kernel-based object tracking," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 25, no. 5, pp. 564-577, Apr. 2003.
[18] P. Zhao, H. Zhu, H. Li, and T. Shibata, "A directional-edge-based real-time object tracking system employing multiple candidate-location generation," IEEE Trans. on Circuits and Systems for Video Technology, vol. 23, no. 3, pp. 503-517, Mar. 2013.
[19] Y. Suzuki and T. Shibata, "Multiple-clue face detection algorithm using edge-based feature vectors," in Proc. IEEE In. Conf. on Acoustics, Speech, and Signal Processing, ICASSP'04, vol. 5, pp. 737-740, May 2004.
[20] B. Babenko, M. Yang, and S. Belongie, "Visual tracking with online multiple instance learning," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, CVPR'09, pp. 983-990, Jun. 2009.
[21] Z. Kalal, J. Matas, and K. Mikolajczyk, "P-N learning: bootstrapping binary classifiers by structural constraints," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, CVPR'10, pp. 49-56, Jun. 2010.
[22] J. Kwon and K. M. Lee, "Visual tracking decomposition," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, CVPR'10, pp. 1269-1276, Jun. 2010.
[23] L. Wang, H. Yan, K. Lv, and C. Pan, "Visual tracking via kernel sparse representation with multikernel fusion," IEEE Trans. on Circuits and Systems for Video Technology, vol. 24, no. 7, pp. 1132-1141, Jul. 2014.
[24] H. Grabner, M. Grabner, and H. Bischof, "Real-time tracking via on-line boosting," in Proc. British Machine Vision Conf., pp. 47-56, May 2006.