Research On ECT Image Reconstruction Algorithm With Approximate Divide And Conquer SVM

Electrical capacitance tomography(ECT)is a kind of technology is mainly used in monitoring the fluid in the pipeline,due to the fluid medium in the pipeline is different,also have different dielectric constants,through applying excitation capacitance voltage measurements on the inner wall of the pipe,and then use the capacitance value for image reconstruction of internal fluid pipeline.This technology has many advantages,so it is widely used.In image reconstruction of ECT system,there is a problem of data measurement error caused by a small number of sensors.There is a "soft field" effect caused by the nonlinearity of capacitance and sensitivity.In order to solve the above problem,the support vector machine(SVM)is used to overcome the nonlinearity,high precision and adaptability.In this paper,the basic principle and system structure of ECT system imaging are introduced.The ECT system model of 12 electrode sensor and ARM+FPGA architecture is given,and KFCMSVM algorithm is used for image reconstruction.During the operation of ECT system,if the data size is too large,the training time is too long or the effect of imaging is not satisfactory.In view of the above problems,an approximate partition SVM(KFCMSVM)algorithm is proposed to improve the above problems.However,the operation of approximate KFCM algorithm will also take a lot of time.So in order to find this effective step,we use the two steps of "divide" and "rule" to approximate the KFCM method.The steps in the global K problem is decomposed into independent sub problems to solve,in the "governance" step,and then one by one to spell out the global solution of the subproblems are obtained,but also that of the solutions of sub problems and how to quickly converge to the global optimal solution.The experiment shows that the use of KFCMSVM algorithm in image reconstruction has a greater advantage than the original SVM algorithm in time efficiency and imaging accuracy.In order to improve the real-time performance of the system,FPGA is mounted on the ARM bus in the way of ARM+FPGA.FPGA is mainly responsible for the operation of KFCMSVM algorithm.ARM is mainly responsible for rebuilding the display and storage results and uploading the results.The KFCMSVM hardware algorithm includes the control module,the Lagrange multiplier module,the offset module and the data classification module.The hardware implementation algorithm greatly improves the speed of image reconstruction.Finally,6 groups of experiments were carried out,in which the accuracy of image reconstruction in single block flow,single drop flow and annular flow was above 95%.