Since the factors such as low signal intensity and complicated environment may cause the spectrum to be cluttered and cause the wave peak to be not obvious, it is difficult for infrared remote sensing to detect precursor chemical vapor accurately. Even the basic digital filtering algorithm is used, the problem can not be solved. For this reason, a filtering algorithm based on Kalman/Wiener superimposion is proposed. By using this algorithm to preprocess spectral signals, the qualitative detection accuracy in subsequent wave peak match can be improved effectively. The experimental data collected are verified on the basis of the improved algorithm. The result shows that compared with other filtering algorithms, the improved algorithm can filter interference signals effectively and can identify characteristic peaks accurately.