In complex and dynamic marine channel environments, the communication capabilities of nodes fluctuate with environmental changes, significantly impacting network communication. Existing channel access algorithms lack the ability to dynamically adapt to network topology and environmental conditions, resulting in substantial degradation in packet loss rate and latency performance under challenging channel conditions. To address this issue, this paper proposes a logarithmic backoff algorithm based on a listening mechanism. The algorithm leverages node listening to dynamically monitor fluctuations in network node density caused by environmental changes. Using a two-dimensional discrete Markov chain model, an optimal contention window is derived as a linear function of the number of neighboring nodes, enabling dynamic backoff strategy adjustment. This approach optimizes the MAC layer access strategy for sensor nodes, significantly reducing communication latency and packet loss rate. Experimental results demonstrate that the proposed algorithm exhibits strong adaptability and robustness in complex channel environments and high-density node scenarios, substantially enhancing the stability and transmission efficiency of communication networks in nuclear radiation environments.