Publicação: Machine learning and image processing to monitor strain and tensile forces with mechanochromic sensors
Carregando...
Data
Orientador
Coorientador
Pós-graduação
Curso de graduação
Título da Revista
ISSN da Revista
Título de Volume
Editor
Tipo
Artigo
Direito de acesso
Resumo
A computer vision (CV) system is proposed for real-time prediction of strain by monitoring the color-changing feature of mechanochromic sensors. Pictures of the sensors subjected to calibration tensile tests were treated with standard image processing methods and analyzed using supervised machine learning (ML) algorithms. Visual strain sensing was demonstrated by linear regression models capable of learning a relation between the applied strain and the reflected structural color. The ElasticNet regression model provided the highest accuracy in the strain prediction task, with a remarkable performance in monitoring real-time strain variation of sensors during a tensile-relaxion cycle. Using calibration curves, the predicted strain can also be employed for estimating the tensile force applied on the mechanochromic sensors. Taken together these results point to potential intelligent systems for noninvasive in-situ visual monitoring of deformations and tensions.
Descrição
Palavras-chave
Computer vision, Image processing, Machine learning, Mechanochromic, Sensors
Idioma
Inglês
Como citar
Expert Systems with Applications, v. 212.
