A soft-sensor for estimating copper quality by image analysis technology

A quasi-line estimation method for copper content based on the color image information of copper alloys is proposed to solve the problem of the time lag and other shortcomings of the off-line chemical analysis method for copper content estimation. First, a 3-CCD color camera was used to obtain the color images of the samples in the standard D65 light source. Then two regression models were developed to estimate the copper contents using the least squares regression (LSR) method. For the first model, the mean values of the R, G and B color channels were chosen as the input variables, while the principal component scores extracted by the principal component analysis (PCA) were used for the second model. Finally, the models were tested by the testing samples for predicting the copper contents. The both mean square errors of the testing samples for the two methods were about 1.5%, which can meet the precision requirements in practice. Experiment results showed that the proposed methods were feasible to quantitatively analyze the copper content in the copper alloy.