Abstract: The world’s largest square kilometer array telescope under construction uses radio interferometry to enhance astronomical data. As the size of the antenna array node increases, the extreme complexity of interference increases squarely. Arrays are deployed in remote mountains and deserts, and existing software-based processing methods struggle to meet energy-efficiency constraints. Aimed at this problem, an efficient cross-correlation calculation acceleration method based on field programmable gate array is proposed. A hierarchical structure was proposed, which could achieve multi-level parallelism in processing quantized data. A structured optimization method for multi-channel segmentation in the temporal window dimension was proposed. A data storage optimization method based on streaming structure was proposed. The test results show that the proposed cross-correlation calculation acceleration method achieves a high throughput rate of 42.76 GFLOPS and a high energy consumption ratio of 7.07 GOP/S/W.