Comparative Analysis Of ET Image Reconstruction Algorithms And Research On Three-dimensional Image Reconstruction

Electrical tomography is an important branch of process tomography technique.Electrical tomography technique is suitable for detecting and controlling the flowingprocess of multiphase flow. It is widely used and also has good prospects in the fieldsof commercial measuring, medical imaging and petrochemical engineering. Alongwith the development of the modern industry, the imaging precision and speed forimage reconstruction in ET technique are required higher and higher. But the existingimage reconstruction algorithms aren’t sufficient to satisfy the requirement ofdetecting the parameters accurately and real-timely, because the ill-conditioning, aswell as under-determinedness, exists in the solution of inverse problem of ET. In orderto get the three-dimensional information of the measured object, thethree-dimensional image reconstruction is absolutely necessary. Therefore, this paperdiscusses the principle and mathematical basis of ET system and does the comparativeanalysis of some familiar image reconstruction algorithms on imaging precision,speed and real-time performance. At last, this paper presents a three-dimensionalimage reconstruction algorithm based on clustering technique.The main contributions and work are as follows:(1) Principles of some familiar image reconstruction algorithms, such asLandweber Iterative method, Newton-Raphson method, Preconditioning LandweberIterative method, Algebra Reconstruction Technique method, Conjugate Gradientmethod, Sensitivity Coefficient method, Linear Back Projection method and TikhonovRegularization method, are described. On this basis, this paper analyses thesecommon image reconstruction algorithms on imaging precision, speed, final values ofobjective function, computational complexity, convergence rate, and real-timeperformance and then summarizes the advantages and disadvantages of each imagereconstruction algorithm. This work lays a foundation for assessing imagereconstruction algorithms.(2) The OPTICS (Ordering Points to Identify the Clustering Structure) clusteringalgorithm is applied to realize the visualization of gray level matrixes from the imagereconstruction algorithms. It offers a new idea for assessing image reconstruction algorithms.(3) This paper proposes a new three-dimensional image reconstruction algorithmbased on clustering technique after analyzing the existing three-dimensional imagereconstruction algorithms. The grid-based clustering algorithm is used to get theclass-centers of two-dimensional images. An image matching algorithm based onfeature is used to match the section images. At last, the three-dimensionalvisualization of the whole system is realized.