Intelligent Intruder Detection System for a Surveillance Rover

Machine learning techniques such as Convolutional Neural Networks (CNNs) and computer vision methods like Deep Neural Network (DNN) models have shown promising results in improving the accuracy and effectiveness of human detection from a surveillance video. This study undertakes a comparative analysis of these two models using the OpenCV library to detect humans in a surveillance footage of an intruder detection (Surveillance) Rover. The proposed process replaces traditional detection methods and involves data collection, preprocessing, and output prediction utilizing neural networks. In the preprocessing stage, human images are gray scaled and illumination is enhanced to improve the accuracy of the prediction. In addition, the DNN model is implemented on a Raspberry Pi. The work aims to develop the adoption of ML and Computer Vision techniques on resource-constrained platforms, providing valuable insights for embedded AI and smart vision applications.