基于神经网络的CT图像重建方法研究

旅客行李隐藏爆炸物CT检测是机场安检的重要部分，图像重建是CT检测技术的重要环节，但实际中往往得不到完全的投影数据。针对不完全投影数据或数据量较少情况的重建问题，提出将神经网络算法应用于图像重建，并利用完全投影条件下FBP算法的重建结果作为网络的目标输出进行权值学习，从而降低网络结构和训练的复杂性。对该算法进行计算机模拟仿真试验，试验结果表明，该算法能有效地完成不完全投影数据的图像重建，且重建图像质量较高。

Research on CT-Based Explosives Detection Based on Neural Network

CT-based explosives detection concealed in passengers luggage is an important portion of security checks at airport. Howerver, the image reconstruction is crucial in detection using computed tomography,but complete projection data are not obtained acturally. Aiming at reconstrction problem of incomplete projection data or fewer data, a neural network approch applied to image reconstruction is proposed in this paper and the network learns the weights between the ideal image that is reconstructed using FBP algorithm at complete projection data and the corresponding projection data, thereby reduce the complexity of structure and training of the network ,and computer simulation are processed.The results show that the proposed algorithm could achieve image reconstruction more effectively with incomplete projection data and the quality of image is improved.