The development and utilization of urban underground space present significant challenges due to complex and unknown geological conditions as well as potential engineering risks. Ground Penetrating Radar (GPR), with its high resolution, adaptability, and non-invasive detection capabilities, has become one of the primary tools for underground space exploration. However, GPR data collected by antennas of different frequencies exhibit distinct advantages and limitations: high-frequency antennas provide higher resolution, which is critical for detecting small-scale or shallow targets, but their detection depth is significantly limited. In contrast, low-frequency antennas offer greater penetration depth, enabling the detection of deeper targets, but at the cost of reduced resolution. This trade-off limits the comprehensive application of GPR technology in urban underground exploration, where both resolution and depth are crucial. To address this limitation, our study proposes a low frequency match filter method. This method first performs spectral analysis on two sets of GPR signals collected by antennas with different frequencies to identify their respective advantageous frequency bands and determine the boundary point between them. The signal collected by the high-frequency antenna is then used as the input, with its high-frequency components remaining unchanged, while the low-frequency components are replaced by the low-frequency signal as the desired output. Based on this, a frequency-segmented matching filter is constructed. Finally, the filter is applied to all radar data collected by the high-frequency antenna, achieving low-frequency compensation for the high-frequency signals. The results from both model experiments and real-world applications demonstrate that the proposed frequency-segmented matching filter significantly broadens the spectrum of high-frequency GPR data, enabling more effective extraction of deep signals while preserving the high resolution of shallow layers. This method greatly enhances the overall capability of GPR in urban underground space detection, addressing the limitations of traditional high-frequency GPR systems with insufficient detection depth in complex geological conditions. In conclusion, the low frequency match filter technology proposed in our study offers an efficient and cost-effective approach for extracting deep signals from GPR, showing promising application potential.