Inspiration The global plastic pollution crisis is threatening marine ecosystems, with millions of tons of plastic waste accumulating in oceans and water bodies. Traditional monitoring methods are limited in scale and efficiency, prompting the need for innovative approaches. The use of satellite imagery, specifically from Sentinel-2, inspired the idea of leveraging cutting-edge technology to detect plastic waste on a global scale, ensuring faster and more comprehensive monitoring of plastic pollution.
What it does This project utilizes Sentinel-2 satellite data to identify plastic waste in ocean bodies. By analyzing spectral signatures that differentiate plastic materials from natural elements in water, the model detects and maps areas where plastics are present, aiding efforts to track, manage, and reduce ocean plastic pollution.
How we built it We built the model by processing multispectral Sentinel-2 satellite imagery, focusing on spectral bands sensitive to plastic materials. After collecting the data, we preprocessed it for noise reduction, cloud masking, and geographic correction. Using machine learning algorithms, we trained the model to recognize the spectral signatures of plastics floating on the water’s surface. The final model was integrated into a pipeline that allows plastic waste detection in various oceanic regions.
Challenges we ran into One major challenge was implementing the model for real-time data analysis. Processing large volumes of satellite data and ensuring timely detection of plastics required optimization in data handling, model computation, and cloud-based infrastructure. Additionally, dealing with environmental variables such as cloud cover, water turbidity, and mixed materials in the imagery posed difficulties in achieving high accuracy in detection.
Accomplishments that we're proud of We’re proud of successfully developing a system capable of detecting plastic waste in oceans with a scalable approach. Achieving accurate detection from satellite imagery, along with our progress in optimizing the model for better performance, represents a significant milestone in environmental monitoring. The project has great potential for positive global impact.
What we learned Through this project, we learned the complexity of working with satellite data, especially in real-time scenarios. We gained valuable insights into image preprocessing, the importance of selecting appropriate spectral bands, and the challenges of building machine learning models for large-scale environmental applications. Additionally, we realized the need for continuous refinement and improvement when handling real-world data.
What's next for Plastic Detection using Satellite Image Analysis The next step is to refine the real-time capabilities of the model, allowing for faster detection and action in response to oceanic plastic pollution. We aim to collaborate with environmental organizations to integrate this system into broader plastic waste management strategies. Expanding the model to detect different types of pollution and applying it to other water bodies and regions around the globe is also a future goal.
Log in or sign up for Devpost to join the conversation.