Ultraviolet (UV) spectroscopy is widely applied in real-time environmental monitoring, especially in diesel vehicle nitrogen monoxide (NO) emissions. However, in field experiments, UV absorption spectrum may exist for different degrees of drifts. Spectral jitters may exist for various reasons such as optical power variation, electrical signal drift, and the refractive index jitters of the optical path for an extended period of time, which causes the detection system to be calibrated. And the pulse xenon lamps as the UV source are characterized by specific emission lines that interfere in spectral analysis directly. For these problems, we proposed the spectral structure matching method based on principal component analysis (PCA), which was compared with the conventional polynomial fitting method to observe feasibility and variability. Further, the UV derivative spectrum was applied to the system appropriately, due to the variation of the absorption peak, and was only related to the target gas by using the above method. We validated our method experimentally by performing the NO UV detection system with the calibration and the comparison test. The results suggested that the calibration relative error was less than 9% and the measurement relative error was less than 6% for this wide range by the proposed processes, which optimized the interference of spectral structures and fluctuation to the system and therefore provided better monitoring. This study may provide an alternative spectral analysis method that is unaffected on the specific emission lines of lamps and is not limited to the spectral region and the target gas.