Position Estimation Of UAV By Image Processing With Neural Networks

Abstract

-This paper presents a study of three artificial neural networks with supervised training and different architectures: network with radial basis function, multilayer perceptron and cellular neural network. These networks were applied to edge detection in aerial and satellite images, for later correlation calculation in spatial domain between these images to simulate the estimation of the geographical position of a unmanned aerial vehicle -UAV. The neural networks results were compared with Sobel and Canny operators.Keywords -Artificial Neural Networks, Image Processing and Air Navigation by Images. IntroductionThe application of UAV stepped up lately due to low operational and manufacture costs compared to conventional aircrafts, don't have crew, among other factors. UAVs are employed in remote sensing, with synthetic aperture radar (SAR) or optical sensors on board, for monitoring cities, human activities or areas of risk such as conflict zones or natural disasters. Radio controlled UAVs usually use antennas with linited reach. In addition, when flying they are subject to possible interferences in the communication process with the control base. Interference in the signal can also occur with UAVs use Global Positioning System (GPS) signals combined with the INS (Inertial Navigation System) for navigation. Possible blocking of GPS signals may also become a limitation to its use. An alternative to overcome these problems is to make use of air navigation systems based real time captured ground information.The use of image processing, computer vision, geoprocessing, and pattern recognition techniques applied to autonomous navigation of UAVs is seen in some studies as in [1], using digital elevation model, in [2] using edge information for the registration of aerial images from satellite image, stereo vision in [3], matching between images in [4], among many.Many papers use ANNs in the navigation system of UAVs, research by [5] shows it. In the literature there are few studies using ANNs in image processing and / or pattern recognition in images in order to help the navigation of UAVs. But the use of ANNs for this purpose can be seen in some studies as in [6], that developed an ANN classify surface in candidate regions for crash in the landing phase of a UAV after the segmentation of the terrain flown. The landing area is chosen using fuzzy rules that use data of the terrain available in the database of the covered areas. The work of [7] also developed an aplication in which ANNs are used in the processing of captured images of the area flown by the UAV. Initially, the texture of the image is extracted by a Gabor filter and an ANN searches the area between some landmarks used for navigation of the UAV, derived from a satellite image. An MLP and a self-organizing map of Kohonen -SOM -were developed for this purpose and their results were compared.

Automatic Position Estimation Based on Lidar × Lidar Data for Autonomous Aerial Navigation in the Amazon Forest Region

“…The paper addresses UAV autonomous navigation over a fully covered and dense vegetation patch of the rain forest on the Amazon region. For this cited context, the procedure of using image processing for UAV positioning, where the peak value of image matching by cross-correlation of segmented images is used, did not work when passive sensors were employed [27,28,30,34], because there are no clear features to be obtained by segmentationsee Figure 1. Therefore, for this situation, we tested the application of an active sensor as an approach to the autonomous navigation problem.…”

“…Many solutions have been proposed for autonomous aerial navigation in GNSS-denied scenarios. One branch of research is to use Computer Vision techniques that allow the UAV to locate itself by means of template matching [24], landmark recognition [25], odometry [26], among others [27][28][29]. Those approaches are usually validated using standard optical RGB cameras.…”

Image Edge Extraction by Artificial Intelligence Schemes for UAV Autonomous Navigation

“…In some of those methods, an image edge detection technique is first applied to both the UAV and georeferenced satellite images; then the UAV position is estimated by comparing the resulting binary images [1,4]. This methodology was applied previously to passive sensors for the visible [2] and infrared thermal bands [3], using a MultiLayer Perceptron (MLP), a type of Neural Network (NN).…”