Suspicious Activity Detection Using Convolution Neural Network and Visual Geometry Group-19

Abstract: Video surveillance plays an important role in today's world. Technology has evolved tremendously as artificial intelligence, machine learning and deep learning become mainstream. Using a combination of the above, there are various systems that helps in distinguishing different types of suspicious behavior from live videos. The most unpredictable thing is the behavior of a person and it is very difficult to find out whether it is suspicious or normal. A deep learning approach is used to detect suspicious or unusual activity in the academic environment and send alert messages to the appropriate authorities if suspicious activity is detected. Surveillance is often done using a series of frames captured from a video. All frames are divided into two parts. In the first part, the features are calculated from the video frames, and in the second part, based on the extracted features, the classifier predicts the class as suspicious or normal.

Keywords: Suspicious Activity, Video Surveillance, Convolutional Neural Network(CNN), Visual Geometry Group(VGG-19)