Application of Deep Learning in the Processing of the Aerospace System's Multispectral Images

Application of Deep Learning in the Processing of the Aerospace System's Multispectral Images

Heorhii Kuchuk (Kharkiv Polytechnic Institute, National Technical University, Ukraine), Andrii Podorozhniak (Kharkiv Polytechnic Institute, National Technical University, Ukraine), Daria Hlavcheva (Kharkiv Polytechnic Institute, National Technical University, Ukraine) and Vladyslav Yaloveha (Kharkiv Polytechnic Institute, National Technical University, Ukraine)
DOI: 10.4018/978-1-7998-1415-3.ch005

Abstract

This chapter uses deep learning neural networks for processing of aerospace system multispectral images. Convolutional and Capsule Neural Network were used for processing multispectral images from satellite Landsat 8, previously processed using spectral indices NDVI, NDWI, PSRI. The authors' approach was applied to wildfire Camp Fire (California, USA). The deep learning neural networks are used to solve the problem of detecting fire hazardous forest areas. Comparison of Convolutional and Capsule Neural Network results was done. The theory of neural networks of deep learning, the theory of recognition of multispectral images, methods of mathematical statistics were used.
Chapter Preview
Top

Background

Vegetation fires are widespread worldwide. They have been documented since prehistoric times, defining the composition and dynamics of ecosystems, including forests and open landscapes. The effects of fires have a significant impact on the environment and society. Fire emissions affect the composition of the atmosphere and the global climate, as well as human health and safety.

One of the largest wildfires happened in California (USA) on July 2018. It was called “Mendocino Complex”. The total area of the fire was almost 460 thousand acres. In less than six months in this state, a wildfire called Camp Fire over 150 thousand acres took place again (The Department of Forestry and Fire Protection of California, 2018). Analyzing US statistics, (Hoover, 2018) compared to the 1980s, now, the number of acres on which fires have increased almost twice, while the number of fires has become smaller.

The primary losses from a large number of wildfires are human life and natural resources. On the other hand, it's worth remembering that wildfires cause considerable economic costs to overcome and eliminate them. For example, in 2017 in the USA, fires injured nearly $20 billion, about $ 2.5 billion was spent in 2008 and 2015, and in 2014, 2010 and 2009, it is less than a billion (Insurance Information Institute, 2017). It is important save human life and health, natural resources, ecology, and economic resources by preventing wildfires.

For monitoring and detection of wildfires terrestrial monitoring, aviation monitoring, space monitoring is being used. Nowadays U.S. Forest Service uses satellite data, laboratories, and research stations to fight wildfires (U. S. Forest Service. Managing Fire, 2019). But such an approach requires people to stay in potentially dangerous forest areas.

To solve the problem neural networks of deep learning will be used. They will be able to recognize fire hazardous forest areas. The multispectral images from the Landsat 8 will be used (U.S. Geological Survey, 2019). The required spectral indices that are able to allocate arid vegetation, moisture content and carbon will be calculated. Such an approach can lead to solve the problem of preventing wildfires.

Key Terms in this Chapter

Index Image: An image is constructed corresponding to the index value in each pixel based on a combination of brightness values in certain channels.

Deep Learning: Is a group of methods that allow multilayer computing models to work with data that has an abstraction hierarchy.

Earth Remote Sensing: Is the method of measuring the properties of objects on the earth's surface, which uses data obtained from aircraft and artificial satellites of the Earth.

Multispectral Image: Is an image which captures data at certain frequencies throughout the electromagnetic spectrum.

An Artificial Neuron Network: Is a computational model based on the structure and functions of biological neural networks.

Complete Chapter List

Search this Book:
Reset