Research On Image Reconstruction Algorithm For Electrical Capacitance Tomography System And Hardware Implementation

The Electrical Capacitance Tomography (ECT) has a lot of advantages ofsimple structure, low cost, non-invasive, fast response, and good safety performance.Research by many scholars, but the characteristics of its own "soft field", as well asthe need to solve the problem of nonlinear shortcomings so that there is still adifficult across the distance for the ECT technology is applied to the actual industry.It also requires a more in-depth research. Support vector machines (SVM) the hotmachine learning, it has reliable, simple in principle, to promote the ability, as wellas good performance in the multi-layered feed-forward neural network performancein the field. It is an excellent method for ECT image reconstruction. There are still alot of problems with SVM algorithm processing of large-scale sample.In this paper, for the problem of training a long time in large-scale sample ofSVM in image reconstruction, the problem of low accuracy, the select block SVMalgorithm is applied to electrical capacitance tomography. If the number of samplessmall than the threshold value of the matrix rows of the certain data, the SVMclassifier is direct used, if the number of samples exceeding the threshold value ofthe row of the data matrix, Choice and Segmentation Support Vector Machines(CSSVM) algorithms is used. For a small unit, select a most suitable combination ofthe small sample from a large sample for this small unit, and training for the bestmodel. Large sample is making into a small sample classification problem with theSVM algorithm; reduce the difficulty of the problem, a shorter time for predicting,higher precision. The experimental results show that the reconstructed image withCSSVM algorithms has a higher classification accuracy and shorter imaging timethan with SVM algorithm alone,.Analysis development processes of FPGA and design methods, in-depth understanding of debugging and simulation process. In the ECT image reconstructionapplications SVM algorithm is achieved with FPGA hardware to accelerate the speedof image reconstruction. Completed the hardware design of the overall architectureof SVM algorithm, and the circuit design of main function modules, and trainingmodule and prediction module are simulated on ISE8.2, select different split unit ofpipeline model for experiments, experiments show that the hardening of SVMalgorithm has high speed image reconstruction.