Turan Göktuğ ALTUNDOĞAN, Mehmet KARAKÖSE

Derin Yapay Sinir Ağları Kullanan Dinamik Bulanık Bilişsel Haritalarla Çoklu Görüde Nesne Takibi

Çoklu görüde nesne takibi, birden fazla görüntüleme aygıtının kullanıldığı görüntüleme sistemlerinde tek bir görüntüleme aygıtından elde edilen görüntü kareleri üzerinde tespit edilen nesnelerin diğer görüntüleme aygıtlarından gelen görüntü kareleri üzerinde de bulunduğu yerin hesaplanmasına dayalı nesne takip işlemidir. Burada bahsi geçen problemin çözümü için genelde görüntüleme sistemi içerisinde bulunan farklı kamera konum ve oryantasyonların içerisinde kullanıldığı hesaplama metotlarından yararlanılmaktadır. Makine öğrenmesi ve yapay zeka tabanlı yöntemlerin bilgisayarlı görü alanında problemlerin çözme kabiliyetinin artmasıyla beraber ÇGNT işlemini gerçekleştirmek için farklı yapay zeka ve makine öğrenmesi tabanlı yöntemlerden yararlanılabilmektedir. Bu çalışmada ÇGNT için bulanık bilişsel haritalardan yararlanan yeni bir yöntem geliştirilmiştir. Bulanık bilişsel haritalar, ele aldığı gerçek dünya sistem veya problemlerine ait özellikleri konsept olarak kabul eder. Daha sonra bu konseptler arasındaki ilişkileri kullanarak iteratif bir şekilde modelleme veya hesaplama işlemini geçekleştirir. Günümüzde endüstri, sağlık, enerji, bilgisayar bilimi vs. gibi birçok alanda problemlerin çözümünde BBH’lar kullanılmaktadır. Bulanık bilişsel haritaların literatürde bilgisayar bilimi alanında sağladığı çözüm önerileri için daha dinamik bir yapıya ihtiyaç duyulmuştur. Bu çalışmada çoklu görüde nesne takibi işlemi için geliştirdiğimiz bulanık bilişsel harita yapısında konsept ilişkilerinin dinamik bir şekilde güncellenmesi için derin yapay sinir ağlarından yararlanılmıştır. Deneysel sonuçların analizi farklı başarım hesaplama işlemleriyle gerçekleştirilmiştir. ÇGNT odaklı yöntemlerin başarım hesaplamasında kullanılan Birleşim Kesişimi (Intersection of Union) yöntemi ile yapılan analizlerde minimum %67,4 maksimum %99,8 ve ortalama %88,2 başarım elde edildiği gözlemlenmiştir. Ele alınan problem için hesaplanan kesişim oranı literatür çalışmaları incelendiğinde çok yüksek bir başarıma sahiptir.

Multi-View Object Tracking with Dynamic Fuzzy Cognitive Maps Using Deep Neural Networks

Multi-view object tracking is an object tracking process based on calculating the location of the detected objects on the image frames obtained from a single imaging device in imaging systems where more than one imaging device is used, also on the image frames from other imaging devices. In order to solve the problem mentioned here, calculation methods in which different camera positions and orientations are generally used in the imaging system are used. With the increase in the ability of machine learning and artificial intelligence-based methods to solve problems in the field of computer vision, different artificial intelligence and machine learning-based methods can be used to perform the MVOT process. In this study, a new method using fuzzy cognitive maps has been developed for MVOT. Fuzzy cognitive maps are graph-based structures that take the features of real world systems or problems they deal with as a concept and perform the modeling or computation process iteratively using the relationships between these concepts. Today industry, health, energy, computer science etc. FCMs are used to solve problems in many areas such as. A more dynamic structure was needed for the proposed methods provided by fuzzy cognitive maps in the field of computer science in the literature. In this study, deep artificial neural networks were used to dynamically update the concept relations in the fuzzy cognitive map structure we developed for multi-view object tracking. The analysis of the experimental results was carried out with different performance calculations. In the analysis performed with the Intersection over Union method, which is used in the performance calculation of the MVOT-focused methods, it was observed that a minimum performance of 67.4%, maximum 99.8% and an average of 88.2% was achieved. When the literature studies are examined, the intersection rate calculated for the problem under consideration has a very high success.

PDF

___

[1] Chen W, Cao L, Huang K. A novel solution for multi-camera object tracking. 2014 IEEE International Conference on Image Processing (ICIP); 2014; pp. 2329-2333.
[2] Jahanshahi P, Masoud A, Moghadam E. Multi-view tracking using Kalman filter and graph cut. 2015 AI & Robotics (IRANOPEN);2015; Qazvin. pp. 1-5.
[3] Yun Y, Gu I, Aghajan H. Maximum-likelihood object tracking from multi-view video by combining homography and epipolar constraints. 2012 Sixth International Conference on Distributed Smart Cameras (ICDSC); 2012; Hong Kong. pp. 1-6.
[4] Chen Z, Liao W, Xu B, Liu H, Li Q, Li H, Yang D. Object Tracking over a Multiple-Camera Network. 2015 IEEE International Conference on Multimedia Big Data; 2015; Beijing. pp. 276-279.
[5] He L, Liu G, Tian G, Zhang, J, Ji, Z. Efficient Multi-View Multi-Target Tracking Using a Distributed Camera Network. IEEE Sensors Journal; 2020; vol 20; no 4: pp. 2056-2063.
[6] Qian Y, Yu L, Liu W, Hauptmann A. ELECTRICITY: An Efficient Multi-camera Vehicle Tracking System for Intelligent City. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW); Seattle; 2020. pp.2511-2519.
[7] Chou Y S, Wang C Y, Chen M C, Lin S D, Liao H Y M. Dynamic Gallery for Real-Time Multi-Target Multi-Camera Tracking. 2019 16th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS); Taipei, Taiwan; 2019. pp.1-8.
[8] Ong J, Vo B T, Vo B N, Kim D Y, Nordholm S. A Bayesian Filter for Multi-view 3D Multi-object Tracking with Occlusion Handling. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2020.
[9] Zhang X, Izquierdo E. Real-Time Multi-Target Multi-Camera Tracking with Spatial-Temporal Information. 2019 IEEE Visual Communications and Image Processing (VCIP); Sydney; Australia; 2019. Pp. 1-4.
[10] Liu X, Dong Y, Deng Z. Deep Highway Multi-Camera Vehicle Re-ID with Tracking Context. 2020 IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference (ITNEC); 2020; Chıngqing; China. pp. 2090- 2093.
[11] Zhou W, Li Z, Gao P. Research on Moving Object Detection and Matching Technology in Multi-Angle Monitoring Video. 2019 IEEE 8th Joint International Information Technology and Artificial Intelligence Conference (ITAIC); 2019; ; Chıngqing; China. pp. 741-744.
[12] Del Rosario J R B, Bandala A A, Dadios E P. Multi-view multi-object tracking in an intelligent transportation system: A literature review. In 2017IEEE 9th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment and Management (HNICEM); 2017; Manila. pp. 1-4.
[13] Gozhyj A, Kalinina I, Gozhyj V. Fuzzy cognitive analysis and modeling of water quality. In Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications (IDAACS);2017; Bucharest. pp. 289-293.
[14] Pajares G, Jesús M. Fuzzy cognitive maps for stereovision matching. Pattern Recognition; 2007; 39(11). pp.2101-2114.
[15] Chen CT, Chiu YT. A study of fuzzy cognitive map model with dynamic adjustment method for the interaction weights, 2016 International Conference on Advanced Materials for Science and Engineering (ICAMSE); 2016; Tainan. Pp.699- 702.
[16] Miao Y. Modelling dynamic causal relationship in fuzzy cognitive maps. 2014 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE); 2014; Beijing. pp.1013-1020.
[17] Mazzuto G, Ciarapica FE, Stylios C, Georgopoulos VC. Fuzzy Cognitive Maps designing through large dataset and experts’ knowledge balancing. 2018 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE); Rio de Janeiro; 2018. pp.1-6.
[18] Bhutani K, Kumar M. Fuzzy inference system & fuzzy cognitive maps based classification. 2015 International Conference on Advances in Computer Engineering and Applications; 2015. pp.305-309.
[19] Georgopoulos VC, Stylios CD. Fuzzy cognitive maps for decision making in triage of non-critical elderly patients. 2017 International Conference on Intelligent Informatics and Biomedical Sciences (ICIIBMS);2018;Okinawa. Pp. 225-228.
[20] Poczeta K, Kubus L, Yastrebov A. Analysis of an evolutionary algorithm for complex fuzzy cognitive map learning based on graph theory metrics and output concepts. Biosystems;179;2019. Pp.39-47.
[21] Altundoğan TG. Derin öğrenme kullanan dinamik bulanık bilişsel haritalar ile gerçek zamanlı nesne takibi. Yüksek Lisans Tezi. Fırat Üniversitesi, Elazığ, 2019.
[22] Altundogan TG, Karakose M. A New Deep Neural Network Based Dynamic Fuzzy Cognitive Map Weight Updating Approach. In 2019 International Artificial Intelligence and Data Processing Symposium (IDAP);2019;Malatya, pp. 1-6.
[23] Altundogan TG, Karakose M. Multiple Object Tracking with Dynamic Fuzzy Cognitive Maps Using Deep Learning. In 2019 International Artificial Intelligence and Data Processing Symposium (IDAP);2019;Malatya, pp. 1-5.
[24] Altundogan TG, Karakose M. Genetic Algorithm Based Fuzzy Cognitive Map Concept Relationship Determination and Sigmoid Configuration. In 2020 IEEE International Symposium on Systems Engineering (ISSE);2020; Vienna pp. 1-5.
[25] Altundoğan, TG, Karaköse M. An Approach for Online Weight Update Using Particle Swarm Optimization in Dynamic Fuzzy Cognitive Maps. In 2018 3rd International Conference on Computer Science and Engineering (UBMK); 2018; Sarajevo (pp. 1-5). IEEE.