Mobile Solution for Alkalinity Estimation from Colorimetric Analysis via CNNs

• Andre Luiz da S. Pereira UFF
• Gustavo L. Silva UFF
• Leandro P. Bernardo UFF
• Luca M. Mendonça UFF
• Paulo H. B. Lopes UFF
• Raphael dos S. Evangelista UFF
• Caroline dos S. Silva UFF
• Helida Vieira UFF
• Mariana Dias UFF
• Renata Recker UFF
• Renato Malbar UFF
• Bruna N. Teixeira UFRJ
• Thiago G. Correia UFRJ
• Aline Machado de Azevedo Novaes UFRJ
• Rafaella Magliano Balbi de Faria UFRJ
• Rogerio Mesquita de Carvalho UFRJ
• Helena C. G. Leitão UFF
• Rossana Mara S. M. Thiré UFRJ
• Maria Luisa A. Gonçalves UFF
• Ana Mehl UFF
• Marco Antonio Gomes Teixeira CENPES
• Leandro A. F. Fernandes UFF

Resumo

We present a computer vision-based solution for estimating alkalinity from images of water samples. Applications include verification of product conformity, industrial process monitoring, and parameter assessment. The method is computationally efficient and implemented as a mobile app. A smartphone camera captures images under controlled conditions using a handheld 3D-printed chamber. Alkalinity is estimated by an artificial neural network using the chrominance distribution of the analyte image, rather than raw RGB data. The approach enables mobility and use outside controlled laboratories. It is suitable as a semi-quantitative analysis method, particularly in critical industries such as oil and gas.

Palavras-chave: Process monitoring, Industries, Graphics, Image color analysis, Oils, Laboratories, Estimation, Computer architecture, Mobile applications, Convolutional neural networks