The invention discloses an intelligent optical active adjustment method based on a
point spread function. The method comprises the following steps: S1, testing and determining a
system misalignment range; s2, acquiring and recording a
point spread function in a current installation and adjustment state; s3, determining ellipticity change of the
point spread function in different view fields according to a
point spread function ellipticity
calculation algorithm; s4, obtaining the current
ellipse rate and the Zernike coefficient of the parameter thereof, and forming an input and output data pair with a corresponding relation with the current misalignment amount; s5, building a
data set composed of the Zernike coefficient and the misalignment amount according to a neural network
algorithm, and obtaining a to-be-trained
data set conforming to standard normal distribution; and S6, building a double-layer
artificial neural network, and determining the range of each hyper-parameter of the neural network according to the sensitivity of each hyper-parameter of the neural network. According to the method for calculating the device misalignment and the
primary mirror deformation by using the focal plane
point spread function, the
system hardware complexity is reduced, the calculation precision of the
system misalignment amount / the
primary mirror deformation amount is improved, and the adjustment process of an optical system is assisted.