Mobile robots lack a driver or a pilot and, thus, should be able to detect obstacles autonomously. This paper reviews various image-based obstacle detection techniques employed by unmanned vehicles such as Unmanned Surface Vehicles (USVs), Unmanned Aerial Vehicles (UAVs), and Micro Aerial Vehicles (MAVs). More than 110 papers from 23 high-impact computer science journals, which were published over the past 20 years, were reviewed. The techniques were divided into monocular and stereo. The former uses a single camera, while the latter makes use of images taken by two synchronised cameras. Monocular obstacle detection methods are discussed in appearance-based, motion-based, depth-based, and expansion-based categories. Monocular obstacle detection approaches have simple, fast, and straightforward computations. Thus, they are more suited for robots like MAVs and compact UAVs, which usually are small and have limited processing power. On the other hand, stereo-based methods use pair(s) of synchronised cameras to generate a real-time 3D map from the surrounding objects to locate the obstacles. Stereo-based approaches have been classified into Inverse Perspective Mapping (IPM)-based and disparity histogram-based methods. Whether aerial or terrestrial, disparity histogram-based methods suffer from common problems: computational complexity, sensitivity to illumination changes, and the need for accurate camera calibration, especially when implemented on small robots. In addition, until recently, both monocular and stereo methods relied on conventional image processing techniques and, thus, did not meet the requirements of real-time applications. Therefore, deep learning networks have been the centre of focus in recent years to develop fast and reliable obstacle detection solutions. However, we observed that despite significant progress, deep learning techniques also face difficulties in complex and unknown environments where objects of varying types and shapes are present. The review suggests that detecting narrow and small, moving obstacles and fast obstacle detection are the most challenging problem to focus on in future studies.

Spoofing of Unmanned Aerial Vehicles (UAV) is generally carried out through spoofing of the UAV’s Global Positioning System (GPS) receiver. This paper presents a vision-based UAV spoofing detection method that utilizes Visual Odometry (VO). This method is independent of the other complementary sensors and any knowledge or archived map and datasets. The proposed method is based on the comparison of relative sub-trajectory of the UAV from VO, with its absolute replica from GPS within a moving window along the flight path. The comparison is done using three dissimilarity measures including (1) Sum of Euclidian Distances between Corresponding Points (SEDCP), (2) angle distance and (3) taxicab distance between the Histogram of Oriented Displacements (HOD) of these sub-trajectories. This method can determine the time and location of UAV spoofing and bounds the drift error of VO. It can be used without any restriction in the usage environment and can be implemented in real-time applications. This method is evaluated on four UAV spoofing scenarios. The results indicate that this method is effective in the detection of UAV spoofing due to the Sophisticated Receiver-Based (SRB) GPS spoofing. This method can detect UAV spoofing in the long-range UAV flights when the changes in UAV flight direction is larger than 3° and in the incremental UAV spoofing with the redirection rate of 1°. Additionally, using SEDCP, the spoofing of the UAV, when there is no redirection and only the velocity of the UAV is changed, can be detected. The results show that SEDCP is more effective in the detection of UAV spoofing and fake GPS positions.

Image matching is one of the most important tasks in Unmanned Arial Vehicles (UAV) photogrammetry applications. The number and distribution of extracted keypoints play an essential role in the reliability and accuracy of image matching and orientation results. Conventional detectors generally produce too many redundant keypoints. In this paper, we study the effect of applying various information content criteria to keypoint selection tasks. For this reason, the quality measures of entropy, spatial saliency and texture coefficient are used to select keypoints extracted using SIFT, SURF, MSER and BRISK operators. Experiments are conducted using several synthetic and real UAV image pairs. Results show that the keypoint selection methods perform differently based on the applied detector and scene type, but in most cases, the precision of the matching results is improved by an average of 15%. In general, it can be said that applying proper keypoint selection techniques can improve the accuracy and efficiency of UAV image matching and orientation results. In addition to the evaluation, a new hybrid keypoint selection is proposed that combines all of the information content criteria discussed in this paper. This new screening method was also compared with those of SIFT, which showed 22% to 40% improvement for the bundle adjustment of UAV images.