The short-term average/long-term average (STA/LTA) method is widely employed in seismic detection and first arrival picking due to its simple principle and calculation. However, the conventional STA/LTA method significantly compromises the algorithm's effectiveness as a result of the time window position, leading to low accuracy in picking and a high false alarm rate. To solve the above problem, improved methods of window position and feature function of the STA/LTA were proposed. Firstly, an improvement approach was adopted based on utilizing the middle window instead of the traditional position effectively mitigated short-term strong interference. Secondly, an enhanced feature function was proposed by incorporating information fusion principles to dynamically adjust weights considering factors such as seismic signal amplitude and frequency changes. This enabled advantageous allocation of weights and enhanced sensitivity towards events. Finally, these two improved methods were combined. Yielding experimental results demonstrate that the improved method can enhance the event detection sensitivity and accuracy in time picking, and effective reduction in false alarm rate.