Uso de inteligência artificial para quantificar e classificar impurezas em cana-de-açúcar para fins industriais

Carregando...
Imagem de Miniatura

Data

2021-03-02

Autores

Santos, Lucas Janoni dos

Título da Revista

ISSN da Revista

Título de Volume

Editor

Universidade Estadual Paulista (Unesp)

Resumo

Artificial neural networks (ANNs) are computational models similar to the structure and functioning of the human brain. They are systems in which basic processing units (neurons) are distributed and interconnected by different connections (synapses), implemented by different mathematical functions, vectors and matrices. They present the ability to acquire and record information from the processing of a database, having a high capacity to map and solve non-linear problems. Due to these characteristics, RNAs were used to quantify and classify the mass percentage of sugarcane and impurities typically found in their harvest, since they can cause increased energy consumption in the preparation of the cane, extraction losses, difficulties in the broth’s treatment and equipment wear. The quantification and classification of the 146 samples were made using digital information from digital images converted into color scales (RGB, HSV, rgb and L), a non-invasive data acquisition method. The classification involved dividing the samples into two classes: 90-100% (class 1) and 41- 89% (class 2) of sugarcane, by weight. The implementation of artificial neural networks were performed in the MATLAB R2018a software with the nnstart tool, using the Levemberg-Marquadt algorithm (trainlm) for the prediction models and the scaled conjugated gradient algorithm (traincsg) for the classification models. The neurons’ number in the intermediate layer was adjusted by trial and error until the best result by the ANNs was achieved. In both processes there were a random division of the samples: 70% for the training of the network, 15% for its validation and the remaining 15% for the ANN test. The results obtained by the ANNs were very promising, since it was possible to predict the amounts of sugarcane, plant material and soil with low absolute errors and high correlation coefficients. It was also possible to classify all samples correctly, even in the face of a complex and multivariate data set such as digital image parameters.
Redes neurais artificiais (RNAs) são modelos computacionais que relembram a estrutura e o funcionamento do cérebro humano. São sistemas em que há distribuição de unidades básicas de processamento (neurônios), interligados por diversas conexões (sinapses), implementadas por funções matemáticas diversas, vetores e matrizes. Apresentam a habilidade de adquirir e registrar informações a partir do processamento de um banco de dados, tendo elevada capacidade de mapear e resolver problemas não- lineares. Em virtude dessas características, foram utilizadas RNAs para quantificar e classificar a porcentagem mássica de cana-de-açúcar e de impurezas tipicamente encontradas em sua colheita, uma vez que elas podem causar aumento no consumo de energia no preparo da cana, perdas de extração, dificuldades para o tratamento do caldo e desgaste de equipamentos. A quantificação e a classificação das 146 amostras foram feitas a partir da utilização de informações numéricas de imagens digitais convertidas em escalas de cores (RGB, HSV, rgb e L), um método de aquisição de dados não invasivo. A classificação envolveu a divisão das amostras em duas classes: 90-100% (classe 1) e 41-89% (classe 2) de cana-de-açúcar, em massa. A implementação das redes neurais artificiais foi realizada no software MATLAB R2018a a partir da ferramenta nnstart, sendo utilizado o algoritmo de Levemberg-Marquadt (trainlm) para os modelos de predição e o algoritmo do gradiente conjugado escalonado (traincsg) para os modelos de classificação. O número de neurônios na camada intermediária foi ajustado por tentativa e erro até alcançar o melhor resultado pelas RNAs. Em ambos os processos houve divisão aleatória das amostras, sendo 70% para a realização do treinamento da rede, 15% para a sua validação e os 15% restantes para o teste das RNAs. Os resultados obtidos pelas RNAs foram bastante promissores, uma vez que foi possível predizer as quantidades de cana-de-açúcar, material vegetal e solo com baixos erros absolutos, além de altos coeficientes de correlação. Foi possível, também, classificar todas as amostras corretamente, mesmo diante de um conjunto de dados complexo e multivariado como o de parâmetros de imagens digitais.

Descrição

Palavras-chave

Artificial Neural Networks, Digital images, Sugarcane, Impurities, Classification, Prediction, Cana-de-açúcar, Impurezas, Classificação, Predição, Redes neurais artificiais, Imagens digitais

Como citar