On June 29th, Professor Sergey Ablameyko, an academician of the National Academy of Sciences of Belarus and the European Academy of Sciences, and a professor at Belarusian State University, visited the Huangdao Forum (Mathematical and Physical Sub-forum) and delivered an academic report titled Analysis of Dynamic Target Movement in Video to teachers and students.
Academician Sergey Ablameyko pointed out that the rapid development of artificial intelligence (AI) in the field of dynamic visual analysis has brought significant benefits to biomedicine and public safety. However, the limitations of traditional methods in analyzing complex motion patterns still restrict the reliability of key decisions. Although the current mainstream target tracking algorithms perform well in simple scenarios, they are difficult to accurately quantify group interaction behaviors (such as aggregation and dispersion), which poses a major risk in cell dynamics research and dense crowd monitoring.
The lecture focused on how to enhance the analytical capabilities of dynamic scenes through a integral optical flow-motion map hybrid architecture, emphasizing the need to ensure that motion patterns are both accurately captured and intuitively understandable to humans. The core of this technology lies in three pillars: spatio-temporal consistency, quantifiable behavior, and cross-scale adaptability. It requires algorithms to reveal the intrinsic laws of motion while providing interpretable decision-making bases through visual forms such as heat maps. Models based on the fusion of optical flow vector integration and convolutional neural networks (CNNs) perform exceptionally well in this regard, presenting motion trajectories in a manner close to physical intuition and significantly enhancing the credibility of medical diagnosis and security early warning.
Academician Sergey Ablameyko mentioned that this technology has demonstrated significant value in practical applications: in the biomedical field, it accelerates the assessment of cancer drug responses through the visualization and modeling of live cell migration trajectories; in video surveillance, it real-time analyzes crowd gathering/dispersal patterns to warn of stampede risks; in smart city management, its optimized dynamic resource scheduling algorithm has increased the emergency response efficiency of transportation hubs by 40%. Additionally, in industrial quality inspection, ecological monitoring, and other fields, this interpretable dynamic visual framework is driving cross-industry intelligent upgrades.
After the report, Academician Sergey Ablameyko engaged in a warm and interactive exchange with the teachers and students on the spot, patiently answering the questions raised by them and offering targeted and instructive suggestions based on his own experience.
Sergey Ablameyko, a professor at Belarusian State University and doctoral supervisor, is an academician of the National Academy of Sciences of Belarus and the European Academy of Sciences, and vice president of the Asia-Pacific Artificial Intelligence Association. He served as the rector of Belarusian State University from 2008 to 2017 and director of the United Nations Institute for Information Studies from 2002 to 2008. He is also an editorial board member of over 10 SCI journals including Pattern Recognition and Image Analysis. Additionally, he is a member of the National Committee for Science and Technology of Belarus and vice chairman of the National Space Committee. His research mainly focuses on computer vision, image analysis, deep learning, and remote sensing information processing. He has led multimillion-dollar space and supercomputing projects in Belarus, published 25 monographs, and authored over 300 SCI papers. His achievements have been applied in medical image diagnosis, intelligent urban management, and video surveillance systems. He has received the highest national science award of Belarus (2002), the Russian Friendship Award (2009), the Gold Medal of the National Academy of Sciences of Belarus (2016), and the Zhejiang Friendship Award (2018).
[Editor:Wanqi Liu]