An Image Reconstruction Method Based on Simulated Annealing and Back Propagation Algorithm for Electrical Capacitance Tomography

摘要

For solving low convergence rate and being vulnerable to fall into a local minimum in BP neural network during training process, a new ECT image reconstruction algorithm based on neural network and Simulated Annealing Algorithm (SAA) is brought forward. In this method, an additive momentum is considered for weight adjustment in BP neural network and adaptive mechanism is adopted to adjust study rate for better performance in convergence of neural network . In addition, the training neural network is guided by SAA to ignore local minimum and is optimized for global minimum. Simulation results show that according to the algorithm based on neural network and SAA, high quality reconstruction images are obtained, which provides a new method for the study of ECT system.