Dredged sludge moisture prediction by textural analysis of the surface image

Texture analyses using the three dimensional color coordinates red (R), green (G), blue (B), and the gray (Y) channels co-occurrence matrix of sludge images were carried out for sludge moisture prediction. Four co-occurrence matrix textural features; information entropy, angular second moment, contrast, and correlation for image texture representation were tested to describe the moisture content variations of the drying sludge. Two of the four features - information entropy and angular second moment, calculated from the sludge image co-occurrence matrixes were modeled by a three, layer perceptron neural network and regression models. The selected models prediction agreed well with the observations with a high correlation coefficient of 0.997 for the neural network and 0.921 for the regression model selected. In the range 350-150% moisture content, information entropy and the angular second moment could present a good description of the texture variation with moisture content. (C), 2000 ISTVS. All rights reserved.