Deep learning and convolutional neural network technologies are increasingly used in the problems of analysis, segmentation, and recognition of objects in images. In this article, a convolutional neural network for automated wildfire detection on high-resolution aerial photos is presented. Two databases of satellite RGB-images with different spatial resolution containing 1457 and 393 high-resolution images, respectively, were prepared for training and testing the neural network. Various techniques of data augmentation are used to enlarge training and test sets generated by data windowing. U-Net neural network with the ResNet34 as encoder was used in research. Neural network training was learning using the NVIDIA DGX-1 supercomputer. An adaptive moment estimation algorithm was used for the optimization of the training process. Special metrics, such as Sorensen-Dice coefficient, precision, recall, F1-score, and IoU value allows measuring the quality of the developed model. The developed algorithm can be successfully applied for early wildland fires detection in practical applications.
- image segmentation ,
- satellite images ,
- forest fire detection ,
- deep learning
- Fires ,
- Image segmentation ,
- Training ,
- Satellites ,
- Databases ,
- Planets ,
- Machine learning
Wildfire is a spontaneous fire that cannot be controlled, occurs outside a special focus, causes material damage, and poses a danger to human life and health. Every year thousands of fires occur for various reasons: seasonally dry periods, thunderstorms, volcanic ignitions. However, last year the human factor has become the main reason for causing irreversible forest fires. Methods of wildfire detection and prevention have varied through the years [1, 2]. Wildfire detection is carried out in five ways: •observation from specially equipped fire observation towers and other structures; •ground observation on foot or by vehicles; •air surveillance using special instruments on aircraft and helicopters; •analysis of information from space; •accounting of messages of locals. The benefit of wildfires is the natural renewal of forests. However, prolonged fires change the composition of the air significantly. The main harm from forest fires is the depletion of flora and fauna, as well as the damage to natural resources. In addition, there is a reason for concern about the harm to human health, in particular to respiratory and circulatory systems. In 2010 the American Heart Association published a scientific statement stating that there is a link between air pollution from tiny particles that appear in the air as a result of wildfires and cardiovascular diseases. There are more than 340 million hectares of forests are annually damaged by fire on the Earth. The largest areas of burning are in Australia and African countries . According to statistics, Russia takes 8th place among all countries of the world which have the most total area of forests destroyed by fire. Thus, there is urgently needed to monitor forests and detect wildfires in the beginning of their spreading. The size of the hotbed of fires makes it possible to detect them from space by means of artificial intelligence methods applied to satellite images [4, 5].
Remote sensing allows to track changes in large areas of the earth’s surface, so the ground services can quickly respond to appeared fires [6, 7]. Computer vision methods are developing rapidly, new papers about object detection are published `every day. But automated image segmentation has not reached to the same quality as the manual marking of high-resolution aerial photos . Manual processing of satellite images shows the highest results, but it takes too much time and human resources. Therefore, the task of satellite image segmentation using computer vision algorithms is particularly relevant.
Image segmentation is a challenging task. Nowadays, various machine learning algorithms for detecting objects on satellite images exist. The main approach of these models is marking image pixels to corresponding classes of objects. This is the task of artificial intelligence. Most existing algorithms for solving segmentation tasks involve using qof deep learning methods, especially convolutional neural networks (CNNs) .
This paper presents a developed convolutional neural network that can be used for wildland fires segmentation. Data preparation, CNN training, testing, and discussion of the achieved results are presented. Our work continues to research, which was presented in [10, 11].
This article presents numerical experiments for developed deep learning algorithms: U-ResNet34. The training and testing process was performed on high-resolution aerial photos of the Resurs and the Planet databases. To implement Numerical experiments there were extracted smaller patches. Training and test sets were enlarged methods using various methods of data augmentation. According to test results for both datasets, U-ResNet34 works satisfactorily.
This work was supported by a grant from the P.G. Demidov Yaroslavl State University NoОП-2Г-09-2019
The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.
KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.
Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.
The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.
FULL Paper PDF file:Wildfire Segmentation on Satellite Images using Deep Learning
Wildfire Segmentation on Satellite Images using Deep Learning
2020 Moscow Workshop on Electronic and Networking Technologies (MWENT), Moscow, Russia, 2020, pp. 1-5,
PDF reference and original file: Click here
Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.