N. K. Priyadharsini and D. Chitra. A Machine Learning and Swarm Intelligence Based Approaches for Anomaly Detection in Extremely Crowded Scenes. Dynamic Systems and Applications 30 (2021) No.8, 1250 – 1272

https://doi.org/10.46719/dsa20213083

ABSTARCT.
The rise in population and versatility of human actions, packed situations have emerged to be hugely prevalent nowadays more than before. Video surveillance is one of the efficient techniques that help to monitor the abnormal activities in real world applications. In the existing work, the accurate detection of abnormal events remains challenging problem in traffic surveillance due to high traffic, occasional background change, overtaking by vehicles, etc. To overcome the above-mentioned issues, in this work, machine learning and swarm intelligence-based strategies are proposed for anomaly detection through the extraction of spatiotemporal features from video clips. In proposed work, an operations act’s set on a video frame to extract the most salient important information. At first, a 2D variance plane is developed for encoding local spatio-temporal changes around every pixel present in the given video. Then the Improved Particle Swarm Optimization (IPSO) algorithm is introduced to separate the most important regions according to mobility information in 2D variance plane. A Grey Level Co-occurrence Matrix (GLCM) is applied on the extracted salient pixels over the video. At last, an Enhanced Artificial Neural Network (EANN) based classifier is utilized for high-level features extraction for anomalous event detection. The result proves that, proposed EANN technique provides superior accuracy, precision, f-measure and recall values rather than the existing methods.

Keywords: Crowded Scenes, Improved Particle Swarm Optimization (IPSO) Algorithm, Grey Level Co-Occurrence Matrix (GLCM), Enhanced Artificial Neural Network (EANN).