Abstract
As the world is continuously growing, the rate of crimes related to trespassing and property invasion are increasing day by day. According to the studies, there are more than 25,000,000 Closed Circuit Television (CCTV) devices are used around the world to monitor the activities in the store premises, organizations, and the properties. These devices provide passive surveillance and does not detect suspicious activity, invasion, and intrusion/trespassing. There are conventional solutions such as the use of edge detection algorithms which are not efficient enough and requires lot of efforts cost and time wise. In recent years, traditional machine learning went through revolution of changes and provided solutions to every sector of economy. The evolution of deep learning has helped in solving great deal of real-life complex problems. The methodology proposed is the approach implemented with the use of deep learning models to identify human intrusion from the footage of surveillance CCTV cameras. The proposed methodology includes a use of Convolutional Neural Network (CNN) with the help of preprocessed surveillance footage from the Virat dataset and the MPII human posture dataset. The proposed methodology also makes use of set of preprocessing steps that includes image cleaning (removing noise) and segmentation, Sobel edge detection, Grayscale conversion, and thermal image conversion using heat maps. The goal of this project is to efficiently identify human intrusion, trespassing, and invasion with the use of common dedicated video surveillance equipment such as CCTV cameras. To achieve this goal, the proposed methodology has used a convolutional neural network which is trained with the help of preprocessed dataset. The proposed system is designed to solve the real- life problem of passive CCTV surveillance.