Deep Convolutional Neural Network For Human Detection And Tracking In FLIR Videos

摘要

A Convolutional Neural Network is used to identify and track humans across video frames. The tracking process is further accentuated by the use of a Kalman filter designed to predict the location of humans in the subsequent video frame. This research is done as part of the SIERRA project at University of Cincinnati, the main objective of which is to provide better situational awareness to the fire-crew during wildland fire situations. This paper focuses on detecting and tracking humans from FLIR video which is a very important aspect of the SIERRA project The results indicate that the CNN and Kalman filter based approach is very effective in achieving these objectives.