Honduran authorities are facing a constant fight against drug trafficking. From 2014 to 2024, they have documented the destruction of 94 coca plantations and 65 marijuana plantations.
However, the use of traditional methods makes these efforts insufficient to completely eliminate illicit crops, which continue to increase in the country, especially in La Mosquitia and the Atlantic coast.
To change the tracking model for these plantations, Honduras is considering the implementation of artificial intelligence technologies (AI).
“What is the idea for implementation is being developed, because this type of technology is not existing in the country yet,” Deputy Inspector Noé Carral, spokesman for the National Drug Enforcement Administration (DNPA), told EL HERALDO Plus.
Although exterminating illegal crops is an important task in the fight against drugs in Honduras – to the extent that there were several years with more operations of this nature than cocaine seizures – it is an activity that involves great resources and very slow execution.
Although plantations are destroyed in some areas, new hectares are often quickly appearing in other areas.
Artificial intelligence could be one of the best allies in this work, as its use could easily be located these plantations and, better still, anticipate where new areas of coca or marijuana cultivation will appear, experts and authorities interviewed agreed.
Difficulties
Crop location work is a task that can last several days, as the help of several experts is needed to demonstrate whether in fact the plantations in those areas are coca or marijuana, since plants can become confused.
Sub-inspector Carral described, for example, that what is a specialist in the field is used, specialists based on the collection of samples that are sent to the laboratories of Forensic Medicine, also specialists in the area to quantify what plants are, coca shrubs or either marijuana, and then, Forensic Medicine is the one who says a final opinion.
Usually, these plantations are usually found in the country’s Atlantic and Mosquitia, as the land is fertile for their growth.
Although reports from the Armed Forces indicate that between 2014 and 2024 94 coca plantations were seized and between 2017 and 2024 at least 65 of marijuana, the Security Secretariat handles more precise figures, but in a different interval of time.
Between 2018 and 2022 alone, 1.7 million coca plants were seized that were on 204.9 hectares of land nationwide, according to the response to the SSSS-1942-2023 request for information from that institution.
65% of these plants were confiscated in Iriona, a municipality of Columbus, the Security report says. In Patuca, in the department of Olancho, they also found 375 thousand coca plants in different operations between 2018 and 2022, which were distributed in 53.5 blocks.
However, this is also an obstacle to being a hard-to-reach region. So far, the authorities have only limited resources for insurance, using aerial and ground surveillance.
On the other hand, one of the advantages of the equipment working in the eradication of these crops is that they already have a mapped area, that is, that the regions where the plants grow dizzily.
How does it work?
AI has the potential to provide accurate and real-time data, improving the efficiency of eradication operations.
Under this model, drones are equipped with high-resolution cameras that can fly over large tracts of terrain or satellite images are used.
At the same time, there is an artificial intelligence program that can detect specific patterns by type of culture, including marijuana and coca leaf, which would be very difficult to identify by land.
Also, there can be programs that allow you to know the exact location of this terrain with coordinates (georrefers), satellite equipment (to obtain real-time images) and anti-drug troops that accompany field reconnaissance operations, experts explained.
To better understand the operation, EL HERALDO Plus spoke with Marco Canales, dean of the Faculty of Engineering of the Central American Technological University (Unitec), and Reymond Bocanegra, head of the career of Engineering in Data Science and Artificial Intelligence at the same university center.
Experts clarified, as a first point, that full training of AI models is required for this specific operation, therefore the work of trained personnel is of the utmost importance to ensure greater effectiveness of the system.
The artificial intelligence model should be trained as if it were a child. In this sense, the satellite makes available a series of images, some of which are usually public; then, in order to reach a specific objective, they must be processed and trained within the AI program.
The training consists of inserting the images to a server and giving instructions, i.e. it is ordered or told the AI what its task will be when focusing the images at different angles and resolutions.
The training process can take six to nine months, properly the collection of images provided by the satellite must be automated so that it starts generating a report in which another person can review it.
It should be noted that to obtain this type of model it is necessary to have several technologies working for the same purpose. These models of deep learning are called neural networks.
Unitec already has a similar program that handles a similar logic to detect this type of illegal plantation, but for another purpose.
This model was created by teachers and students with the intention of detecting places to put solar panels. The model is the same and, as engineer Bocanegra explains, it is only necessary to train AI in the specific report.
The Deputy Minister of Security, Julissa Villanueva, for her part, explained that they could also be used in other areas, such as network and data tracking, money laundering and discovering all those financial resources that move between banks.
Expensive investment
The implementation of artificial intelligence technologies in the detection of drug plantations implies a rather expensive investment.
Acquiring and maintaining AI-equipped drones, high-resolution cameras and specialized software can come at a very high cost.
Adding to this, you need the training of all the staff who would make use of the equipment.
However, the implementation of AI could reduce time, resources, money and precision, according to Nelson Castañeda, director of Security and Justice of the Association for a Fairer Society (ASJ).
In addition, Castañeda proposed that artificial intelligence can support predicting the most likely routes used by these criminal networks, which come from the south of the continent and go north, because obviously that comes to benefit quite a lot in the investigation.
Although it is considered very expensive, state secretariats already work with artificial intelligence, including the Forest Conservation Institute (ICF), which uses remote sensors and satellite equipment to detect changes in forest cover, allowing a distinction between forest and non-forest areas.
The problem is that the institution’s current technology does not have enough capacity to detect illicit crops in the country, said Manuel Erazo, head of the ICF Forest Heritage Center.
In addition, to install this model it is necessary to measure certain confidence intervals. It is important to manage the security and privacy challenges prudently… there are always going to be false detections very low at the beginning, so it is very important that this information is not going to be taken as a type of accusation against people who also have nothing to do with an illicit process, warned the engineer Canales, of Unitec.
In the end, Canales valued that artificial intelligence has to be seen as a tool… important to handle it with the same care that one would handle a tool, let’s say, that can be double-edged.
To effectively counter drug trafficking, Honduras will have to form alliances with institutions that already manage artificial intelligence, adapting this technology for the location of illicit plantations.
This article has been translated after first appearing in El Heraldo