Neural networks are usually used in combination with classical restoration methods, which are used as parameter identifier and restoration filter in restoration. At present, the commonly used neural network in restoration technology is based on Hopfield, which can achieve fast and stable convergence, and the restoration accuracy is also very high. Neural network has strong parallel operation ability, nonlinear mapping ability and adaptive ability, especially suitable for describing complex nonlinear systems. Wu et al. Designed a restoration algorithm based on neural network. This method can not only restore the image, but also protect the edge of the image, which is better than the traditional Laplacian algorithm. Wang et al. Studied the neural network restoration algorithm based on pattern learning. The algorithm is robust to noise cancellation, and its parallel processing improves the real-time performance of the algorithm. Wong et al. Implemented an adaptive regularization restoration algorithm without specifying parameters in advance by using neural network method. According to different noise types and spatial distribution, the network automatically adjusts parameters. Bao et al. Use multilayer perceptual model neural network to restore the image and realize the regularization of edge protection. This method uses subband coding and artificial neural network to perceive image parameters, effectively eliminates noise, and is suitable for image restoration with high contrast. Experiments show that the algorithm can protect edge information and has strong robustness. When talevski et al. Studied the nonlinear restoration problem, they derived a general nonlinear model, established the mapping function from degraded image space to real image space, and implemented the algorithm with neural network. Experimental results show that the method has good dynamic characteristics, can correct nonlinear distortion, and is robust to random noise. Many excellent characteristics of neural network make it very suitable for application in restoration algorithm, which has been considered as one of the best restoration methods by scholars for many times. Celebi et al. Studied the method of extending Lyapunov equation from discrete time domain to continuous time domain neural network. The results show that continuous time domain neural network is one of the best methods for image restoration. Sun used Hopfield based neural network in the measurement and tracking of sea features, and thought that neural network is the most suitable and one of the best ways to develop sea feature recovery measurement system.