A New Vision-Based and GPS-Signal-Independent Approach in Jamming Detection and UAV Absolute Positioning Assessment

The Unmanned Aerial Vehicles (UAV) positioning in the outdoor environment is usually done by the Global Positioning System (GPS). Due to the low power of the GPS signal at the earth surface, its performance disrupted in the contaminated environments with the jamming attacks. The UAV positioning and its accuracy using GPS will be degraded in the jamming attacks. A positioning error about tens of kilometers off the true location was reported in the research articles, due to jamming. Assessing the performance accuracy of UAV’s GPS receiver during the jamming attacks has great importance in maintaining the UAV’s flight safety, especially in the military applications. Currently, the developed and presented methods in jamming detection and mitigation relies on the analysis of the GPS signals. On the other side, many study works are being performed for effective and successful jamming attacks for security, military and intelligence purposes. To detect the occurrence of errors in the GPS positioning, therefore, it is necessary to research and introduce some new independent techniques of the GPS signal processing methods. The presented method in this paper to detect the occurrence of the jamming attacks and to assess the UAV's GPS performance is an independent method from the GPS signals processing. This method is an image-based approach which uses the UAV images taken from its flight path. As the camera is a passive sensor, its performance will not be affected by the external signals and electronic warfare such as jamming attacks. The UAV's flight trajectory is extracted using visual odometry (VO) from the UAV images, and then this trajectory is compared with the UAV's trajectory derived from the corresponding GPS data. In this method, there is no need for the initialization of the coordinate system parameters of VO as it is a relative method in agent positioning. Also, to reduce the drift error in VO only a few images within a selection window is used in the computation of the relative and absolute trajectories of UAV. For reduction of drift-error of visual odometry within this selection window, sliding widow bundle adjustment technique is used here. This process repeated for each UAV position from the begin to the end of the UAV trajectory. For each selection window, two trajectory descriptors correspond to the UAV trajectory from VO and its replica from GPS data are computed.  Then these trajectories compared using their trajectory descriptors. Two trajectory descriptors were defined here to compare these trajectories: The Normalized Distance of Consecutive Points (NDCP) trajectory descriptor and the Consecutive Directions Angle (CDA) trajectory descriptor. These trajectory descriptors are the functions of the UAV’s positions in its flight path within each selection window. The comparison of these trajectory descriptors is done using the cosine similarity index (CSI). CSI takes into account the normalized angle between two codimension normalized vectors to compare them. If there is a problem in the UAV's GPS performance, the cosine similarity index will differ from its ideal number one or 100%. The results show that the proposed method, using these trajectory descriptors, can detect the occurrence of jamming attacks and to assess the accuracy of the UAV's GPS performance during its flight.