It is important to improve the reliability of the eddy current non-destructive testing. It is generally difficult to reduce the probability of non-detection and the number of false alarms same time. Model Based Measurement (MBM), including estimation of state, failure diagnosis and trend analysis, has excellent results on general estimation. Actual results of MBM vary with the application domain, mathematical model and data processing. The time series model is a description of system in time domain based on equivalent output. The Kalman filter is an efficient method for suppressing the disturbance and improving the state estimation. A new method, called as eddy current testing (ECT) using a model based measurement, is presented. Two time series models are used for the normal status and the abnormal status in ECT. The Kalman filter on these models is applied to give useful information for decision making. Results of the experiment on aluminum plates demonstrate that this method is useful to improve the detection reliability.