The traditional pulse clustering method is based on DSP and GPU processing platforms. In the face of scenes with large signal bandwidth, pulse collision, and high data throughput, the real-time performance of the DSP-based method, which runs in sequence, cannot be satisfied. The GPU parallel execution method, which is efficient but higher power consumption, does not meet the requirements of lightweight and low-power design for spaceborne radar processing systems. In order to improve the efficiency of pulse clustering method and reduce the power consumption of radar processing systems, this paper proposes a highly parallel two-dimensional real-time clustering structure based on FPGA. By using the pipeline processing characteristics of FPGA, this method can match the sampling rate of AD and it can be applied to the large bandwidth collision pulse scenes with the high requirement of real-time clustering. The effectiveness of this method is demonstrated by the results of typical broadband collision pulse measurement scenes. It has the characteristics of stronger real-time performance and wider applicability.