Title:
Unauthorized Person Detection using Thermal Imaging and Gait Recognition for Intra-Building Security
Author:
Samit Ari, Mohammad Iman Junaid, Narayan Prasad Sharma and Irshad Ali
Abstract:
The present disclosure relates to surveillance systems to monitor suspicious activities, for protecting private areas and restricted premises from unauthorized entry. The disclosure provides thermal imaging, which eases the segmentation of human silhouettes from the background with significantly less IR emission. The presently disclosed gait recognition based on multiple cameras is used for the detection of unauthorized access and automatic location monitoring of unauthorized persons through an intelligent network of cameras with mutual communication and coordination. The data acquisition system is of present invention is based on multiple thermal cameras, Here the first camera (C1) detects the person whether he is authorized or not to enter into the building using the primary database, the unauthorized person is allowed to walk through the second camera (C2) where he is registered in the stranger database, and the third camera (C3) is placed at the exit of the building which uses both primary and stranger database for detection. The gait model (Gait Energy Image) with timestamp of each person is created while registering them to the database. Deep learning technique like convolution neural network (CNN) is used to recognize the thermal gait sequences.