Computer vision for bamboo control in the Miño and Sil river basins 

The Confederación Hidrográfica del Miño-Sil (CHMinoSil) and the Computer Vision Center initiated a collaboration to develop and implement technology to monitor the persistent growth of bamboo in the Miño and Sil rivers in Galicia. 

Bamboo: an invasive specie 

Originally from Asia, bamboo has proliferated in various parts of the world due to its ornamental and practical uses. However, its spread has become a significant environmental issue, particularly in ecosystems where it is detrimental. The Miño and Sil river basins in Galicia are among the areas affected by this invasive species. 

Negative consequences

The spread of bamboo has several adverse effects, including:   

• Competition with native species 
• Changes in ecosystem structure 
• Invasion problems 
• Infrastructure risks 
• Challenges in management and control 

These issues were the primary objectives for CHMinoSil, leading them to collaborate with the CVC. They aimed to explore whether the latest AI and computer vision techniques, combined with information that can be extracted from the Plan Nacional de Ortografía Aerea (PNOA) images, could offer a viable solution. 

From the perspective of the CHMinoSil, Javier Mosquera, Head of Technical Section, explained that the process “demanded huge investment of time and resources, considering the region covers a territory of 17.619km²". He noted the positive impacts of the project: “More than 600 sites with various bamboo species have been geo-referenced, enabling effective planning of management and control measures. This information serves as a basis for a future request to include potentially invasive genera in the Spanish Catalogue of Invasive Alien Species”. 

How can computer vision tackle the problem? 

The Computer Vision Center’s proposal is based on advanced processing techniques that extract detailed information and automate tasks, facilitating effective environmental conservation management. In this case, the CVC analysed PNOA images to generate a dataset and trained a model to automatically identify bamboo. 

Coen Antens, head of the CVC Innovation Unit, and Alberto Rubio, Research Engineer, explain the solution in the video attached below. The project’s results demonstrated the feasibility of using aerial pictures combined with computer vision and deep learning techniques to detect bamboo in the Miño and Sil river basins. One key conclusion was the accurate detection of bamboo zones from satellite images. Additionally, the project emphasized the importance of continuous information gathering to implement effective control and management measures, ensuring the maintenance of biodiversity in the affected ecosystems.

A new path for detecting other invasive species 

The project showcased the potential to adapt and apply this technology to other areas and invasive species, such as cane, which also has a significant presence in Spain. Coen Antens explains that the next steps will involve further exploring these applications: "We have planned to continue collaborating with CHMinoSil in the detection of cane and also apply it to Catalonia, where we also have this problem." 

Mosquera also added that the CHMinoSil expects to continue collaborating with the CVC to apply these techniques for managing other invasive species affecting the demarcation.