Reconstruction Algorithms Based on Artificial Neural Network for Electrical Capacitance Tomography

For the non-linear characteristic and "soft filed" feature of electrical capacitance tomography system, in this paper, based on the analysis of neural network basic algorithm principles, optimizes parameters in the activation function and error function for BP network are optimized. The simulation experiment shows that the learning speed and convergence rate of the BP network has been improved. It brings many excellent capacities for the new reconstruction algorithm such as fast image reconstructing, easy system implementing and strong noise resisting etc, and it supplies a new idea for electrical capacitance tomography image reconstructing.