With the development of modern industrial applications, laser processing requirements with complex processing environments and objects, large dynamic range, high efficiency, and high precision are becoming more and more urgent. And the on-line monitoring and real-time optimization of laser processing parameters is an important solution. At the same time, a variety of optical signals and changes in surface optical properties can be generated during the interaction between laser and material. They are closely related to processing parameters, processes, and target properties. Therefore, spectral measurements of corresponding optical signals could reveal the machining process and status, indicating an important on-line monitoring means of laser processing. Spectral measurements with the characteristics of high resolution and rich spectral information have been used for almost all laser processing processes, including laser welding, laser cutting and drilling, laser cleaning and polishing, micro-nano structure preparation, and additive manufacturing. This paper analyzes and summarizes the spectral measurement techniques, including plasma spectroscopy, reflection spectroscopy, and nonlinear optical spectroscopy, applied in on-line monitoring of laser machining. Based on the spectral measurement of plasma signals excited during single and multiple-pulse processing, the qualitative and quantitative monitoring of chemical composition during laser processing can be achieved. In addition, laser focus can be adjusted in real-time according to the relative intensity variances of characteristic peaks. The laser processing processes related to thermal effect can also be monitored and regulated based upon the plasma temperature. As a non-destructive monitoring method over a relatively long distance, reflection spectroscopy can effectively monitor the cleanliness, damages, chroma, and compositional changes of the material surface by measuring the integral spectral power of the reflected light signal a specific band, position and intensity of characteristic spectral peaks and bands. The nonlinear optical signals excited under certain conditions, including the harmonic signals, the fluorescence signals, and the Raman signals, can also provide additional methods of spectral measurements. Although their application scenarios are limited, they provide a new monitoring method for component analysis, focus, and material damage. Furthermore, the future development trend of spectral measurement, including the collaborative monitoring of multiple optical signals, the combined monitoring of spectral, acoustical, temperature and image signals in on-line monitoring of laser processing, has prospected. Meanwhile, the combination of artificial intelligence technology, on-line monitoring and laser processing will further promote the intelligent development of laser processing technology.