Study On Statistical Process Control Methods For Monitoring Image Data

With the widespread use of machine vision systems in industrial applications,more and more product information is presented in the form of image data.How these images can be utilized within the framework of statistical process control has become a new promising area.This study focused on the monitoring of images with uniformity or a specific pattern and presented approaches to solve the problem of unknown shape/number of fault and high correlation between the intensity values of neighboring pixels.This study first proposed a new control charting method based on the Exponentially Weighted Moving Average(EWMA)statistic and region growing technology to monitor images with unknown shape of a fault.More details about the implementation of this monitoring scheme are discussed.Performance of the proposed method is compared with that of an alternative method proposed by Megahed et al.via simulations.It is shown that the proposed method is more effective in detecting the emergence of a fault and accurate in estimating the size/location of the fault.This study also proposed a new control charting method based on the sum of the generalized likelihood ratios to monitor images with unknown number of faults.More details about the implementation of this monitoring scheme are also given.Performance of the proposed method is analyzed under both single and multiple faults scenarios via simulations.Comparison with the method proposed by Megahed et al.under the single fault scenario shows that the proposed method is more effective in detecting the emergence of the fault.Considering the high correlation between the intensity values of neighboring pixels,this study proposed the multivariate generalized likelihood ratio based control charting method.More details about the implementation of this monitoring scheme are also list out.Performance of the proposed method is analyzed under both single and multiple faults scenarios via simulations.Comparison with the method proposed by Megahed et al.under the single fault scenario shows that the proposed method is more efficient in detecting the emergence of the fault.At last,the multivariate generalized likelihood ratio based control chart is applied into two experiments to highlight how practitioners can implement and make use of this control charting method in image monitoring applications.Results indicate that the proposed control charting method can not only rapidly detect the emergence of the faults but also provide good estimates of the occurring time and locations of the faults,which can help the practitioners identify the assignable cause and recovery the process as soon as possible.