In order to improve the image reconstructed quality affected by soft filed feature and the speed of dynamic\r
on-line data processing in Electrical Resistance Tomography, we propose a fast image reconstruction\r
algorithm based on H\u221e filtering theory. Mainly, on the H\u221e filtering principle, a dynamic system is\r
formulated firstly, whose inputs have unknown disturbances including noise errors and model errors, and\r
the outputs have the estimation errors. Then, making the H\u221e norm of this dynamic system as a cost\r
function, a fast H\u221e filtering algorithm is proposed whose criterion is to guarantee that the worst-cast\r
effect of disturbance on estimation error is smaller than a given boundary. Experimental work was carried\r
out for three typical flow distributions. Results showed that H\u221e filter method improves the resolution of\r
the reconstructed images and gains the strong robustness and anti-interference performance in unknown\r
interference noise conditions. In addition, it dramatically reduces the computational time compared with\r
the traditional Gauss-Newton iterative and Kalman filter methods. Therefore, the method is suitable\r
for on-line multiphase flow measurement.
