Noisy Laser Wavelength Classification Based on Self-weighting Low-rank Multi-view Spectral Clustering

Laser is widely used in machining and measurement fields. Detecting the wavelength variation of laser under the interference of noisy light sources is of great significance for its application. In this research, a multi view spectral clustering method used on Fresnel grating diffraction image is proposed to address the issue of laser wavelength variation measurement. Firstly, the theoretical practicability of using static Fresnel grating diffraction image for laser wavelength stability detection is analyzed. Then, diffraction image features from multiple views are extracted. Finally, a self-weighting low-rank multi-view spectral clustering (SL-MVSC) method is proposed to classify the obtained multi-view grating diffraction image data matrix to achieve detection of laser wavelength variation. In the optimization process, there are two variables involved, one of which can obtain a closed form solution, and the optimization of the other variable is transformed into solving a quadratic programming problem. The clustering performance of this method is tested in various noise interference environments. The experiment results show that this method performs better on several clustering performance metrices compared with other four methods. Meanwhile, the t-SNE results indicate that this method can distinguish samples from different clusters. The study proves the practicability and innovation of proposed multi-view method for detecting laser wavelength variation in static Fresnel grating diffraction image.