Anomaly detection in crowd scenes via cross trajectories

AKPULAT M., EKİNCİ M.

Applied Intelligence, cilt.55, sa.6, 2025 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 55 Sayı: 6
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1007/s10489-025-06338-z
  • Dergi Adı: Applied Intelligence
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, PASCAL, ABI/INFORM, Applied Science & Technology Source, Compendex, Computer & Applied Sciences, Educational research abstracts (ERA), INSPEC, Library, Information Science & Technology Abstracts (LISTA), zbMATH
  • Anahtar Kelimeler: Abnormal event detection, Finite time braid entropy, Local and global anomaly, Trajectory-based detection
  • Karadeniz Teknik Üniversitesi Adresli: Evet

Özet

This paper proposes a novel finite-time braid entropy (FTBE) theorem to extract feature vectors to detect abnormal events occurring globally and locally in crowds. Detecting abnormal events or behavior in crowd movements is a key research topic regarding community security and management. A trajectory- based method depending on the FTBE theorem and the distribution of motion vectors is presented to determine abnormal events. The FTBE theory determines the complexity of the pattern occurring during the movement of the trajectories describing the behavior. In most studies in the literature, the image is divided into equal regions and the solution is produced by separating every behavior into more than one zone. However, this may result in incorrect results. Our study separated the behavior within a certain time interval into location-independent motion clusters. Each cluster indicated a behavior, which was represented by a feature vector derived from the distribution of FTBE and motion vectors. The learning model and fully connected deep neural network were used to detect which cluster was behaving abnormally in the local area. In addition, abnormal events were determined globally by the step braid entropy score (SBES) value calculated for the current scene. The method was tested using the UMN, UCSD and UCF-Crime databases. The experimental results of the method showed an alternative approach to the detection of abnormal behavior.