基于SA-BP算法的ECT图像重建机制的研究与实现

In this paper, an Electrical Capacitance Tomography (ECT) image reconstruction mechanism based on SA-BP algorithm is presented to avoid the disadvantages that slow convergence of BP neural network and reaching local minimum easily during the training. Firstly, adding a momentum in the weights adjustment formula and regulating learning rate by using adaptive method, the convergence rate is improved. Secondly, when neural network reach local minimum, the introduction of simulated annealing method can guide the neural network to jump out of local minimum and find the global minimum. Simulation experimental results show that the SA-BP algorithm can provide high quality results and this new algorithm presents a feasible and effective way to research on image reconstruction for ECT.