When using low-frequency ultrasonic TOF technology to measure the thickness of thick multi-layer composite media, there is a tendency for valid echoes and near surface interference waves to overlap and be difficult to separate between ultrasonic reflection echoes, which affects the measurement accuracy of TOF. In response to such problems, this article proposes an adaptive matched filtering processing method based on L1 norm. Based on the spatiotemporal consistency of near-surface interference, effective echoes exhibit distinct transition characteristics. In the spatiotemporal domain, the collected ultrasound array signals are adaptively matched to achieve the separation of effective echoes and near surface interference waves, thereby enhancing the resolution of effective echoes. The simulation data processing has demonstrated the effectiveness of the proposed method, with a 3dB increase in signal-to-noise ratio. The experimental results show that this method can efficiently solve the problem of overlapping ultrasonic signal separation, and has shown broad application prospects in thickness measurement scenarios of solid rocket engines.