Object Detection in Aerial Photos Using Neural Networks

The paper describes the method of objects detection on aerial photographs using neural networks. The aim of this paper is to present an object detection algorithm by neural networks using sliding window. Main idea and main benefit of this concept is that image processing by sliding window with different sizes and that user can set possibility threshold for neural network classifying. This approach can be used with any neural network types because the goal of neural network in the method is to classify current part of image. For our experiments we took convolutional neural network and aerial photos. Also in this paper described the extension of this method. It’s an algorithm that allows post processing of data obtained as a result of the operation of neural networks. The problem of searching for aircraft in images is considered as an example. Some results of aircraft detection are presented in this paper. Image processing took place in distributed data processing system that also described in this paper. © Springer Nature Switzerland AG 2020