Space optical communication has been rapidly developed in recent years because of its advantages, which include a high transmission rate, strong anti-interference, good confidentiality, and small equipment sizes. Although space optical communication has many advantages, there are also atmospheric factors, such as atmospheric scattering attenuation, turbulence flicker, and turbulence speckle, that cause various degrees of interference to lasers. Hence, the quality of the light spot received by the receiving end of the communication system will degrade, and the detection accuracy of the light spot will also be affected. Therefore, to enhance the stability of optical communication systems in the presence of atmospheric interference, it is imperative to explore the spot localization algorithm.

A grid-based neural network localization algorithm for spot centers is proposed to improve the stability of optical communication systems under atmospheric interference. First, the region of interest (ROI) is extracted from the light spot detected by coarse tracking, and the image is segmented by the maximum entropy thresholding method. Then, the segmented image is divided into grid cells, and the effective response area of each grid cell is separately calculated. Finally, the sequence of the effective response area of each grid is inputted into a pre-trained BP neural network, and the accurate coordinates of the light spot center position are predicted.

To verify the effectiveness of the algorithm proposed in this study, the centroid algorithm, circle-fitting algorithm, and proposed algorithm were used to perform comparative experiments on image spot positioning errors for two sets of images with different types of light spots. The absolute errors of two different types of light spots are shown in Figs. 7 and 8. The results of the absolute errors show that the absolute errors of the proposed algorithm are the smallest. The maximum error is associated with the circle-fitting algorithm, and this is followed by the centroid algorithm. The experimental results of the root mean square values of the absolute error are shown in Table 1. The average value of the root mean square values of the absolute error in the two sets of images using the proposed algorithm is only 70.11% of that of the centroid algorithm and only 52.27% of that of the circle-fitting algorithm. In the experimental results of the absolute error standard deviation in Table 2, the average value of the absolute error standard deviation in the two sets of images using the proposed algorithm is only 77.80% of that of the centroid algorithm and 65.73% of that of the circle-fitting algorithm. Compared with the traditional centroid algorithm and circle-fitting algorithm, the absolute error of the proposed algorithm is minimal, and the spot location accuracy is better.

This study proposes a method for the accurate location of the center of a light spot in space optical communication systems. The maximum entropy threshold image segmentation method is used to segment the ROI of a beacon light spot detected by coarse positioning, and the image is then divided into 2×2 grids. Next, the effective response area of each grid cell is computed and inputted into a pre-trained BP neural network to get the accurate coordinates of the spot center position. The experimental results show that the absolute error of the proposed algorithm is minimal and that the center position of the light spot can be accurately located. The proposed algorithm can effectively inhibit the influence of atmospheric interference on the positioning accuracy of the center of the spot and provide an effective guarantee for the stability of the space optical communication system. Hence, the proposed algorithm has significant practical application value.