To improve the imaging resolution and detection accuracy of ground penetrating radar(GPR) methods in complex surrounding rock conditions in tunnels, a reverse time migration approach is proposed combined with an adaptive attenuation compensation model. This approach involves deriving the instantaneous amplitude for each sample from the collected data profile by extracting the instantaneous amplitude spectrum and its attributes within each time window. The method uses these time-dependent instantaneous amplitude attributes, which reflect the signal′s attenuation along the profile to construct an adaptive linear exponential polynomial that matches the observed attenuation characteristics. This results in an analytical signal attenuation model that is used for compensation. By applying this attenuation-compensated model to reverse time migration imaging, it is aimed to maximize the retention of deep anomalous detail in lossy media, thus improving the potential for high-fidelity imaging of structural defects. The feasibility and practicality of the method are evaluated using a synthetic invert arch model and field experiments conducted on tunnel bottoms and linings. The results indicate: (1) The diffraction superposition and multiple reflection regions exhibit the highest extremes in instantaneous amplitude attributes. (2) Multiple exponent adaptive compensation can objectively characterize radar signal attenuation and ensure high-fidelity profile characteristics. (3) Data with a high signal-noise ratio(SNR) can be compensated by using n≥5 order, while data with a low SNR should be compensated by using n≥9 order.