Moving target tracking and recognition method for unmanned airborne surveillance systems

Abstract

This paper considers vision-based multiple moving target-tracking and target-type recognition methods for unmanned airborne surveillance systems. The detection of moving objects and target-type recognition in a moving image frame are the essential parts of airborne surveillance systems. We propose an optical flow-based object detection method with image stabilization functions to detect moving objects in a moving image frame, and a combination of the Support Vector Machine (SVM) with Convolutional Neural Networks (CNNs) model for the target-type recognition. The experiment of an airborne surveillance scenario using a quadcopter with a camera is conducted to demonstrate the performance of the proposed method. © ICROS 2017.