Convolutional Neural Network for Measurement of Suspended Solids and Turbidity

The great potential of the convolutional neural networks (CNNs) provides novel and alternative ways to monitor important parameters with high accuracy. In this study, we developed a soft sensor model for dynamic processes based on CNN measurement suspended solids turbidity from single image liquid sample be measured by using commercial smartphone camera (Android or IOS system) light-emitting diode (LED) illumination. For this, an dataset samples illuminated white, red, green, blue LED light was taken used train fit multiple linear regression (MLR) different color lighting, evaluated which gives more accurate information about concentration particles in sample. We implemented pre-trained AlexNet model, MLR estimate total (TSS), values particles. proposed technique obtained goodness (R2 = 0.99). best performance achieved white light, accuracy 98.24% 97.20% TSS turbidity, respectively, operational range 0–800 mgL−1, 0–306 NTU. This system designed aquaculture environments tested both fish feed paprika. motivates further research aquatic such as river water, domestic industrial wastewater, potable among others.