Sergey Ablameyko is Academician of National Academy of Sciences of Belarus and Academician of the European Academy, Fellows of IEE, IAPR, NAS, BEA, IAIPT, AE, SRAD and AAIA. Currently, he is a professor of Belarusian State University. He has published more than over 450 scientific papers, 15 books.

Title: People and crowd behavior identification in video

Abstract:
People and crowd motion analysis is an important task in many applications and it is widely used in video surveillance systems to prevent many undesirable events and cases.
We show a formalization of the problem of detection and tracking of people and crowd in video. At first, we defined person, group of persons and crowd motion detection types and formalized them. For crowd, we defined three main types of its motion: direct motion, aggregation and dispersion. Then, we defined crowd behavior parameters and especially crowd abnormal behavior detection features. Based on these formalizations, we developed algorithms for detection and tracking people and crowd in video sequences for indoor and outdoor environment.
A novel method based on integral optical flow will be shown to analyze crowd motion and identify aforesaid crowd behaviors for videos obtained by stationary cameras in public places. We accumulate basic optical flows to form integral optical flow and use it to separate background and foreground and obtain intensive motion regions. Based on information extracted from integral optical flow, we analyze pixel motions statistically for each frame to obtain quantity of pixels moving toward or away from each position and their comprehensive motion at each position.
We then define and compute regional motion indicators to describe motions at region-level. Thresholds for motion intensity, quantity and motion direction of pixels are used together to segment regional motion maps and identify crowd behaviors.
The speech also presents the formalization of the detection and tracking of people of abnormal behavior in video sequences. The criteria characterizing: the quality of detection of accompanied objects, the accuracy of determining the location of the object on the frame, the trajectory of movement, the accuracy of tracking a variety of objects are considered. Based on the considered generalization, algorithms have been developed for detecting abnormal behavior and the type of people using tracking through detection and convolutional neural networks to detect people and form signs. Neural network features are included in a composite descriptor, which also contains geometric and color features to describe each detected person in the frame.
Experimental results will be shown that confirm that our method can identify and locate the crowd behaviors successfully.