By Thiago Nunes Kehl, Viviane Todt, Maurício Roberto Veronez, Silvio Cesar Cazella
The most desirable target of the current learn used to be the advance of a device to observe day-by-day deforestation within the Amazon rainforest, utilizing satellite tv for pc pictures from the MODIS/TERRA sensor and synthetic Neural Networks. The constructed device presents parameterization of the configuration for the neural community education to allow us to choose the simplest neural structure to deal with the matter. The software uses confusion matrices to figure out the measure of good fortune of the community. A spectrum-temporal research of the learn sector used to be performed on fifty seven photos from may well 20 to July 15, 2003 utilizing the educated neural community. The research enabled verification of caliber of the carried out neural community category and likewise aided in figuring out the dynamics of deforestation within the Amazon rainforest, thereby highlighting the titanic capability of neural networks for photograph type. notwithstanding, the complicated job of detection of predatory activities first and foremost, i.e., new release of constant alarms, rather than fake alarms has no longer been solved but. therefore, the current article presents a theoretical foundation and elaboration of useful use of neural networks and satellite tv for pc photos to strive against unlawful deforestation.
Read Online or Download Real time deforestation detection using ANN and Satellite images: The Amazon Rainforest study case PDF
Best remote sensing & gis books
A concise, self-contained monograph on laser distant sensing and its purposes, this article discusses the ways that lasers can be utilized to remotely degree the ambience and the hydrosphere. It offers a old viewpoint and stories the elemental physics had to comprehend the topic.
Clever seonsors are revolutionizing the area of process layout in every little thing from activities autos to meeting traces. those new sensors have skills that go away their predecessors within the dirt! They not just degree parameters successfully and accurately, yet in addition they manage to increase and interupt these measurements, thereby transformng uncooked info into really worthwhile details.
City distant Sensing is designed for higher point undergraduates, graduates, researchers and practitioners, and has a transparent concentrate on the advance of distant sensing know-how for tracking, synthesis and modeling within the city surroundings. It covers 4 significant components: using high-resolution satellite tv for pc imagery or substitute resources of photo date (such as high-resolution SAR and LIDAR) for city function extraction; the advance of more suitable snapshot processing algorithms and methods for deriving exact and constant details on city attributes from distant sensor info; the improvement of analytical innovations and techniques for deriving symptoms of socioeconomic and environmental stipulations that be triumphant inside of city panorama; and the improvement of distant sensing and spatial analytical innovations for city development simulation and predictive modeling.
- Applied Hydrogeology of Fractured Rocks: Second Edition
- Measuring Precipitation from Space: EURAINSAT and the Future (Advances in Global Change Research)
- Remote Sensing of Land Use and Land Cover: Principles and Applications
- Handbook of Research on E-Planning: ICTs for Urban Development and Monitoring
- Principles of Synthetic Aperture Radar Imaging: A System Simulation Approach
Additional info for Real time deforestation detection using ANN and Satellite images: The Amazon Rainforest study case
The difficulty of classification of spectrally similar classes were also observed by Todt et al.  and Bischof et al. , who performed a comparative study of statistical techniques and ANNs, checking the difficulty of classification in both methods. In order to demonstrate the difficulty of separating two classes of similar spectral signature, the same network with exactly the same parameters withdrew all points relating to the savannah class from the training samples was trained. Three trainings were carried out and all of them converged for the expected MSE before the iteration number 500 due to the extinction of confusion between classes.
The implementation of this work was carried out using the programming language Java, using AWT (Abstract Windowing Toolkit) and Swing components for creating the graphical interface. The Encog Framework  was incorporated to develop the neural network module and the database management system MySQL Server  was used for storing data related to the processed images. 2 Development Tool The neural deforestation detection tool was developed based on the methodology demonstrated in reference , in order to detect daily deforestation from MODIS/ TERRA images of the study area.
With the use of the Java programming language and Encog Framework, a neural module was implemented, using a Multilayer Perceptron neural network. This module made it possible to train neural networks and verify its generalization ability for the sets of tests. We opted for the free parameterization of the neural network developed through the GUI tool in order to add flexibility to the software created. The data from images of the area of study relating to tile H11V09 of MODIS/ TERRA sensor was used in the development process.