ANTIPARA (Analysis of Tiny Particles in Aquatic Environment): A Water Scanning Device for Microplastics

Microplastics particles have become an important ecological problem due to a huge amount of plastics debris that ends up in the sea.It is expected that the number of affected marine species will rise as research on this topic increases.Deep learning shows a potential for solving complex problems without the need for a physical understanding of the underlying system, and hence offers an elegant solution.The application of convolutional neural networks for the identification of microparticles in different aquatic environment was demonstrated.The measurements were carried out in real-time using a Raspberry Pi, a digital microscopic camera, and neural network computation, hence demonstrating a portable and low-cost environmental aquatic sensor.Phyton programming language was used to encode the input in the raspberry pi which serve as the brain of the device.The network model is trained using 1000 datasets where 70% was designed for training and validation, and 30% was for testing.The deep learning approach produced a good performance with 97.65% -88.32% -84.00% trainingvalidation-testing accuracy for the Convolutional Neural Network model.The actual field tests conducted, showed a high percentage accuracy (90%) on different aquatic environment when compared with laboratory tests using Ultraviolet Visible Spectroscopy.The average discharging time was 1.18 hour which denotes that the scanning device can be used for a long period of time.The program and platform of the scanning device were functional.It can be concluded that the water sensing device is a great help in keeping our water clean and safe.