Abstract: To address the issue of processing gain degradation in matched filtering caused by spectral distortion of linear frequency modulation (LFM) long-pulse acoustic waveforms propagating through time-varying underwater acoustic channels, this paper proposes a sub-pulse coherent integration method with matched filtering. The long-pulse waveform is partitioned into M sub-pulses for segmented matched filtering. After aligning the matched filter peaks of these sub-pulses, the inter-pulse phase differences are accurately estimated and compensated. The phase-compensated and aligned peaks are then coherently integrated to restore the processing gain. Simulation and experimental results demonstrate that the proposed method effectively overcomes channel coherence limitations through sub-pulse partitioning, significantly mitigating processing gain degradation in long-pulse matched filtering caused by spectral distortion. By preserving the advantageous large time-bandwidth product of long-pulse waveforms, this approach provides a novel framework for addressing active detection challenges of LFM long-pulse signals in time-varying underwater acoustic channels.