| Border Surveillance and security are of prime concern for any country. The utilization of modern smart technologies empowers strong border security. It is an imperative need to advance these technologies for better security. In this thesis, we are employing Machine Learning techniques in Remote Video Surveillance for real-time threat level detection and classification of the targets, crossing borders. This will ultimately enhance the performance of Remote Video Surveillance (RVS) system deployed at or near borders. As a result, we can achieve military decision support system and can utilize the resources efficiently for effective Border Surveillance.;The algorithm used for the machine learning based detection of objects in the videos in this research is Viola-Jones algorithm. The algorithm requires a training set of both positive and negative images, for the purpose of which, a collection of positive and negative images was used for the training of the algorithm for objects such as humans, vehicles, and handguns.;A threat level classifier and alert warning system were also added to classify and annotate the videos in real-time for each frame. The threat level classifier performs four-fold categorization of the real-time video into -- safe, low, medium, and high (danger). The alert warning system specifies the type of warning based on the type of intrusion (human, vehicle, or weapon) detected. For the algorithm proposed in this work, the accuracy for the human detection is an average of iv 94.93%, the accuracy for the vehicle detection is an average 95.2%, and the accuracy for the weapons detection is an average of 97.67%. The accuracy of our proposed method (97%) was much higher than that of the previously published method (64%) for object detection.;The results of this research showed that the proposed machine learning based real-time processing of Remote Video Surveillance (RVS) is a practical and feasible method for accurate intrusion and threat detection, which is very useful for real-time processing of remote surveillance video at banks, ATMs, borders, and security checkpoints for public safety, security, and protection. |
