An efficient abnormal behavior detection and tracking using deep learning in real time applications
Loading...
Date
item.page.authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
With today s enhancing requirement for security, most importantly
newlinein public regions such as crowded streets, schools, supermarkets, train
newlinestations, and airports, surveillance cameras are employed for observing day
newlineto-day activities and recognizing suspicious events. Abnormal behavior
newlineidentification from huge video data is of high importance for smart video
newlinesurveillance that has become a famous experimental topic in modern years.
newlineHowever, it is still highly complex to recognize suspicious behavior because
newlineof the critical situation in the video frames and the suspicious behavior s
newlinedeceptive nature. Numerous techniques have been utilized to recognize the
newlineabnormal behavior in the modern days. Multiple techniques concentrate on
newlinethe movement trajectories to differentiate the suspicious activities. A behavior
newlineis considered abnormal if the specific trajectory varies from the learned
newlinepatterns. However, due to the tracking functionality, the techniques are
newlinecomplex for evaluating complex and crowded scenes. In order to address the
newlineocclusion and complex situations, numerous methodologies extract the
newlinealternative attributes from the motion cue and low-level appearance. But, the
newlineattributes employed by these approaches are manually crafting that have a
newlinelimitation to the anomaly identification process, since it requires previous
newlineknowledge and complex to draw out the complex video information.
newlineNowadays, deep learning frameworks have been utilized successfully to
newlinerecognize suspicious behaviors. The primary reason why the technique is
newlineeffective in anomaly identification is where the deep frameworks can
newlineunderstand high more significant features from the raw images. Subsequently,
newlinethe abnormal activity identification based on the deep learning strategies
newlineminimizes the computational complexity.
newline