Deep Learning-Based Person Detection and Classification for Far Field Video Surveillance

This paper presents a deep learning-based approach to detect and classify persons in video data captured from distances of several miles via a high-power lens video camera. For detection, a set of computationally efficient image processing steps are considered to identify moving areas that contain a person. These areas are then passed onto a convolutional neural network classifier whose convolutional layers consist of the GoogleNet transfer learning. Despite the challenges associated with the video dataset examined in terms of the low resolution of persons appearing in the video data and the presence of heat haze and camera shaking, the developed approach generated 90% classification accuracy.