针对传统CCD标定中由于修正像差造成标定算法复杂、精度不高、结果不稳定的特点,提出了用4层前向神经网络模型修正CCD像差的方法。为了检验算法,用修正像差后的CCD对空间点进行三维重建,并与传统方法作了对比。结果表明,基于神经网络模型的CCD像差修正方法可以获得比传统方法更高的精度和稳定性。 Aiming at the characteristics of traditional CCD calibration, such as the calibration algorithm is complex, the precision is not high and the result is unstable, the method of correcting the CCD aberration with four layers forward neural network model is proposed. In order to test the algorithm, the space point was reconstructed by using CCD with corrected aberration and compared with the traditional method. The results show that the method of CCD aberration correction based on neural network model can obtain higher accuracy and stability than the traditional method.