Expanding Small UAV Capabilities with ANN: A Case Study for Urban Areas Inspection

Abstract

Aims: Autonomous Unmanned Aerial Vehicles (UAVs) provide an effective aerial alternative for urban areas inspection due to its cost and safety when compared to more traditional methods. The purpose of this paper is to report the development of a system capable of analyzing digital images of the ground and of detecting potential invasion, unauthorized alterations on the ground and deforestation in protected natural areas. Study Design: The project was developed in collaboration between researchers in the context of the master's program in Science and Technology in Computation of the Federal University of Itajuba. Place and Duration of Study: Institute of Mathematics and Computation and Institute of Advanced Studies, between March 2012 and July 2013. Methodology: The Images are captured by a camera mounted on an autonomous electrical helicopter, which overflies the area under inspection. For the processing of the images an artificial neural network technique called Kohonen SOM (Self Organizing Map) will be used. The processing is actually composed of a sequence of steps that seek to collate the final common characteristics of a given image. Results: The Kohonen SOM allows grouping the pixels of an image with similar characteristics. In the case of this work, the pixels become widespread in two classes - white and black. After processing, there is a new output image with rearranged colors is produced. The same process can be used for detecting flaws in transmission lines in all three spectrums mentioned in this article. Conclusion: Today UAVs are already being used in many fields today and will certainly be largely used for urban areas surveillance. The use of the helicopter for land inspection the land showed significant results especially considering the low vibration level produced by its electric motor.