Objective Space laser communication technology combines the advantages of fast laser communication speed, wide bandwidth range, good confidentiality, and flexible application in wireless communication. It has gradually been widely utilized and has become a major research hotspot. Optical communication video transmission is undoubtedly an important application scenario for space laser communication. Traditional video transmission methods can deal with cumbersome image data, especially in scenarios such as inter- satellite and satellite- to-ground communications, where data acquisition is challenging. This often causes problems in data transmission and storage, adding considerable stress to storage units. Compressive sensing technology, which combines sampling and compression, bypasses the Nyquist sampling theorem, significantly reducing data in the link and alleviating pressure on the transmission channel. Although the current traditional block compressive sensing (BCS) algorithm improves the processing speed of compression reconstruction, it applies a unified sampling rate to each block, despite the different image information contained in different blocks. When the image content is divided into target and background, then the current processing mechanism typically under-samples the target and over-samples the background, leading to low data utilization and suboptimal reconstructed image quality. Therefore, we need to consider the status of different image blocks and further optimize the algorithm. Methods We focus on space laser communication video transmission. It uses the image centroid as the judgment feature value, calculates the centroid error between frames, and evaluates the changing speed of the image block. This approach helps determine the sampling rate for the current frame image and generate a measurement matrix by reducing the sampled data for blocks with high inter- frame correlation and increasing the sampling rate for blocks with low inter-frame correlation and overall data utilization. Then, we use FPGA as the main control chip to build an experimental system for video image compressive sensing transmission and reconstruction. The system tests the video image transmission under spatial light, comparing the reconstructed image results between the proposed algorithm and the traditional algorithm. Results and Discussions We simulate this algorithm based on a set of natural scene video extraction image sequences, setting the total sampling rate to 0.1. Each frame of the image is compressed and reconstructed using this algorithm. At the same time, a comparative experiment is conducted with the traditional BCS algorithm at the same sampling rate, comparing the corresponding reconstruction results of different frame images (Fig. 3). At low sampling rates, other algorithms produce reconstruction results with significant random noise and blur, affecting image quality. However, the proposed algorithm achieves good reconstruction and restoration of image details. To further evaluate the algorithm's performance, we use some typical metrics such as the image peak signal-to- noise ratio (PSNR), structural similarity (SSIM), normalized root mean square error ( NMSE), and gradient magnitude similarity deviation (GMSD). Under multiple groups of specified sampling rates, the average values of the multi- frame reconstructed image data indicators are compared ( Figs. 4 and 5). The proposed algorithm outperformed others in various performance indicators at different sampling rates. Especially when the sampling rate is extremely low, the traditional typical measurement matrix can hardly reconstruct the original image, while our algorithm can basically retain the characteristics of the original image. Taking the sampling rate of 0.1 as an example, the average PSNR value is about 8 dB higher, and the overall average SSIM is more than 9% higher than that of other algorithms. We develop a spatial optical video transceiver board based on FPGA chips and build two spatial optical video transmission principle terminals. Using a communication rate of 1.25 Gbit/ s, we use frame-by-frame transmission to collect 1550 nm wavelength optical video stream signals and sample a total of 200 frames of video image sequences as data for the transmission experiment. In our experiments, the receiver collects compressed image sequences for reconstruction and solution, and further combines all reconstructed images to obtain a video (Fig. 8). At a 0.2 sampling rate, the PSNR of reconstructed video images by our algorithm is generally higher than 35 dB, which is generally more than 5 dB higher than that of other algorithms. At the same time, SSIM indicators have also improved by more than 8% compared with other algorithms. Conclusions We propose a method for video compression transmission in space laser communication systems by optimizing the traditional compressive sensing algorithm through the comparison of centroid differences between video image frames. The image block sampling rate and measurement matrix are designed based on the distance between each image block and the center of mass and the degree of image change between frames. This method improves the sampling efficiency of compressive sensing to a certain extent and reduces the impact of block oversampling and undersampling on image reconstruction quality in traditional sampling schemes. At the same time, we built an algorithm principle hardware testing system based on FPGA to provide a guarantee for the experimental verification of the algorithm. Experimental results show that compared with the traditional block compressive sensing algorithm, the proposed algorithm reconstructs video results with better quality, particularly at low sampling rates, providing better reconstruction effects for each frame of image in the video stream, which has certain practical value.
