This UGR application will make it easier to identify the origin of the wood at customs

He illegal timber trade It is an activity that is not only harmful to the treasury, but also to the environment. To control it, the University of Granada has developed a application that allows to identify the origin of the wooda, thanks to an artificial intelligence system, which contributes to reducing the illegal traffic of these goods.

This GO IMAI application has been presented in Madrid.

This tool has been developed by the Wood Identification and Artificial Intelligence Operational Group (GO IMAI) and designed for security forces and agents who control the illegal traffic of wood. This application allows you to establish a early warning when they are faced with a shipment that presents doubts about the species of wood.

The results of the project have been presented this morning at the Royal Academy of Engineering, in Madrid, by Elsa Enríquez Alcalde, deputy general director of Forest Policy and Fight against Desertification of the Ministry for the Ecological Transition and the Demographic Challenge; Luis García Esteban, director of the ETSI Montes Forestal and Natural Environment; Francisco Herrera Triguero, Professor of Computer Science and Artificial Intelligence at the University of Granada; Alberto Romero Cagigal, Secretary General of the Spanish Association of Wood Trade and Industry, and Jesús Gálvez Pantoja, Lieutenant Colonel and Head of the Central Operating Unit for the Environment, of the SEPRONA-Civil Guard Headquarters. During the event, Rosana Montes Soldado, professor in the Department of Computer Languages ​​and Systems at the University of Granada, gave a demonstration of the developed application.

The University of Granada, the Polytechnic University of Madrid and the Spanish Association of Wood Trade and Industry (AEIM) make up GO IMAI, the results of which represent a notable advance in the transparency of international wood trade. The IMAI Operational Group began work in May 2021 and is co-financed by the Ministry of Agriculture, Fisheries and Food and the European Agricultural Fund for Rural Development (FEADER), within the Call for Innovation Projects 2020, in which it obtained the highest evaluation of all the initiatives presented.

What is the origin of this application

The project has its origin in the need on the part of customs agents, inspectors of the Administration and the bodies and security forces of the Stateespecially the Nature Protection Service (SEPRONA) of the Civil Guard, to have a tool to establish an early warning against suspicious shipments of wood from illegal trade.

Currently, the identification of wood only at a macroscopic level is not possible and requires the intervention of highly specialized personnel for a guaranteed identification and expert use. Given this, the combination of knowledge in macroscopic anatomy of wood and artificial intelligence has allowed GO IMAI researchers to work on the design and implementation of a tool that facilitates this identification by the agents themselves in a simple and quick.

The project also seeks to respond to other social challenges, contributing to the conservation of forests, their biodiversity and thereby contributing to mitigating the effects of climate change. And it is that the environmental effects of illegal logging include deforestation and loss of biodiversity. The World Bank estimates that governments around the world lose between $10 and $15 billion each year from illegal logging.

Likewise, estimates from the Intergovernmental Panel on Climate Change indicate that global deforestation negatively influences climate change, since it accounts for 15% to 20% of global greenhouse gas emissions.

This situation underscores the need for prevent the trade of wood at origin and destination without proof of its legality. In this sense, the mobile application developed allows the authorities to have a free resource to control compliance with the regulations on international timber trade, both with regard to the European Timber Regulation (EUTR), which will be replaced by the European Regulation against Deforestation (EUDR), as in relation to the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES). Likewise, it will also be an excellent tool for professionals in the sector, researchers and hobbyists.

Almost two years of work

The results are the fruit of almost two years of joint work and the experience of researchers in the fields of wood anatomy and artificial intelligence. GO IMAI has developed an intelligent technological solution that, in a matter of seconds and from a photograph taken with a magnifying lens attached to a mobile phone, recognizes a species of wood with a high success rate.

During the process, the work team has documented and photographed 400 species of wood, from various mobile cameras and magnifying lenses, determining the configuration of the most useful information. Next, the problem domain has been analyzed and a methodology has been proposed that optimizes the quality preprocessing of the macroscopic images of the wood, which are taken for the training of the intelligent classification system after segmentation processes and increased data.

Subsequently, the model based on deep artificial neural networks or ‘deep learning’ has been optimized to allow its execution on the device itself during the classification from a sample image, without resorting to cloud computing, thus saving costs. costs of communication with servers.

The two applications developed are native and free, one of them for distribution in the Apple Store and running on iPhone, and the second for distribution on Google Play and running on Android.

Likewise, the project has included two usability tests to obtain information on the ease of use of these two applications, considering accessibility aspects.

GO IMAI has implemented a web database with which the complete information of the 400 species has been documented and which is consulted from the apps thanks to the application of more than 40 different filters, which allows access to the file of a species not just as a result of the AI ​​model prediction. Once published, the app can be used by customs agents and state security forces.

GOIMAI in figures

The intelligent system developed by GO IMAI has worked with a high volume of macroscopic images, 19,340 in total, with a resolution of 4,000×3,000 pixels., from the sampling of 531 species, using 5 different magnifying lenses. In the months of experimentation, the project has generated and discarded 50 different models, which could have involved some 600 hours of training in a high-computing system.

The model has considered an average of 12 raw samples per class, and in an advanced optimization phase, another 28 models have been generated, totaling 280 hours of training.

The calculations that are made in the training (applying scripts with different objectives) modify the original data in various ways to obtain a data increase (5 steps) and quality improvement (3 steps). These mods perform grayscale changes, equalization, brightness, contrast, and saturation changes, rotation, cropping, and scaling.

The programming has been approached using the languages ​​of Python and Kotlin, and those of the development of the apps on iOS and Android.

Used bookstores:

  • Tensorflow 2.9.1
  • Tensorflow addons 0.15.
  • Panda 1.4.0
  • Numpy 1.23.5
  • Scikit-learn 1.0.2
  • Pillow 9.0.1
  • open cv 4.5.5
  • And up to 300 packages installed as dependencies

Deixe um comentário

O seu endereço de e-mail não será publicado. Campos obrigatórios são marcados com *