Behrooz Moradi; AbasAli Mehraban; Morteza Mohammadi
Abstract
Autonomous landing is a key challenging in the domain of UAV navigation systems. Developing an autonomous landing system requires a precise estimation of the UAV pose relative to landing marker, particularly in vision systems this involves precise Helipad recognition. It seems that traditional approaches ...
Read More
Autonomous landing is a key challenging in the domain of UAV navigation systems. Developing an autonomous landing system requires a precise estimation of the UAV pose relative to landing marker, particularly in vision systems this involves precise Helipad recognition. It seems that traditional approaches including cascade classifiers, image matching and segmentation techniques to have major challenges in different weather conditions and scales. On the other hand, convolutional neural networks (CNNs) have been introduced as a powerful tool in the visual recognition systems in the recent years but the high computational cost of this techniques, limited their performance in the low cost and light weight UAVs. The aim of this paper is to compare the convolutional neural networks and cascade classifier for helipad detection. The results show that CNNs are invariant under translation, rotation, scaling and occlusion. The detection accuracy of this method is 99.1 % which is 3 % more than cascade classifier while its running time is suitable for real time UAV applications.
Behrooz Moradi; mohammad javad valadan zoej; mojtaba jannati; somayeh yavari
Abstract
In the absence of satellite ephemeris data and inner geometry of satellite’s sensor, utilization of Rational Function Models (RFMs) is one of the best approaches to georeferencing satellite images and extracting spatial information from them. However, since RFMs have high number of coefficients, ...
Read More
In the absence of satellite ephemeris data and inner geometry of satellite’s sensor, utilization of Rational Function Models (RFMs) is one of the best approaches to georeferencing satellite images and extracting spatial information from them. However, since RFMs have high number of coefficients, then usually high number of control points is needed for their estimation. In the other hand, RFM terms are uninterpretable and all of them causes over-parametrization error which count as the most important weakness of the terrain-dependent RFMs. Utilization of optimization algorithms is one of the best approaches to eliminate these weaknesses. Therefore, various optimization algorithms have been used to discover the optimal composition of RFM’s terms. Since the mechanism of these algorithms is different, the performance and feature characteristics of these algorithms differ in the discovery of the optimal composition train-dependent RFM’s terms. But the existing differences not comprehensively analyzed. In this paper, in order to comprehensive assessment the abilities of Genetic Optimization Algorithm (GA), Genetic modified Algorithm (GM), and a modified Particle Swarm Optimization (PSO) in terms of accuracy, quickness, number of control points required, and reliability of results, are evaluated. These methods are evaluated using for different datasets including a GeoEye-1, an IKONOS-2, a SPOT-3-1A, and a SPOT-3-1B satellite images. In terms of accuracy achieved, difference between these methods was less than 0.4 pixel. In terms of speed of evaluation of parameters, GM was 10 to 12 time more quickly in comparison with two other algorithms. In terms of control points required, degree of freedom of modified PSO was 45.25 percent and 27 percent more than GM and GA respectively, and finally in terms of reliability, the dispersion of RMSE obtained in 10 runs of three algorithms are relatively same. These results indicated that accuracy and reliability of all three methods are almost the same, speed of GM is higher and modified PSO needs less control points to optimize terrain-dependent RFM
