Intrusion Monitoring in Military Surveillance Applications using Wireless Sensor Networks (WSNs) with Deep Learning for Multiple Object Detection and Tracking

Abstract

Terrestrial Wireless Sensor Networks (WSNs) are used in military environments for region surveillance, healthcare systems for soldiers, and, smart transport, and logistics, etc. In surveillance applications, the sensor nodes are deployed randomly in the field to observe the events of interest, movement of humans, or vehicles. In these sensor networks, the image or video is captured by the camera module. Many times it becomes difficult to correctly detect the intrusion or anomalous activity in the field because the image being captured maybe not clear enough due to prevailing weather conditions, the amount of light, and other reasons. In this paper, in addition to a WSN Surveillance System for military applications, we have used Convolutional Neural Network (CNN) for analyzing and understanding the content of the captured images and videos. CNN is a deep learning neural network that detects and tracks automatically the important features without any human supervision. The distinctive layers of each class are learned by themselves and have the highest accuracy of prediction. The results of the implementation for four test images captured in different conditions show an accuracy of 92%. The results of the video tracking yield the Object Tracking Efficiency of 80.35%.