Energy Parameter Prediction Methods Of Variable Energy Imaging

X-ray multi-energy imaging technology is proposed to solve the problem that complexcomponents’ information incomplete under the conventional detection, which is owing to thelimitation of the imaging system’s dynamic range. Multi-energy imaging mainly acquires theimage sequence in fixed tube voltage interval. When detect an unknown structure object, theenergy parameters can not match the effective thickness, leading to low collection efficiency.To solve this problem, the prediction method is proposed. This prediction algorithm is able tomatch the energy parameter with the object’s thickness variation, so the system can acquirethe structure information very quickly.In this paper, two algorithms are proposed to predict the parameters. The first algorithmbuilds a physical imaging system based on exponential model, which contains two functions.The first one is about the image gray and tube voltage of effective thickness. The second onebuilds the relationship between tube voltage and gray difference of the effective thickness andadjacent thickness. By solving the two functions we can predict the best tube voltage foradjacent thickness in X-ray graded energy imaging. Exponential forecasting model is basedon nonlinear fitting to establish a relationship between tube voltage and gray difference. Fromthe perspective of information optimization, a new prediction method is proposed. It based onkalman filter algorithm. On the basis of system model established previously, it uses kalmanfilter to build time update equations of image gray, and state update equations of effectivethickness. Then use the two equations to estimate image gray of adjacent thickness, andestablish the relationship between tube voltage and image gray. On this basis, predict the bestimaging tube voltage. Finally, the paper use standard steel block that thickness changescontinuously and engine cover whose thickness changes irregular, then perform theverification tests respectively. Make the accuracy over95%as a standard. Comparative testresults show that the two prediction methods: for varying continuously thickness object,kalman filter prediction method’s average accuracy is more3.94%than the exponential modelapproach, but for complex structure, the prediction method based on exponential model has a higher prediction accuracy about0.15%. On the whole point, forecasting methods based onkalman filter has a higher accuracy.