It is significant for archaeology to detect fragments of ceramics on-line and rapidly. Laser-induced breakdown spectroscopy (LIBS) technique is well suited for the classification detection of ceramic fragments online for its characteristics such as non-contact, fast, and no sample preparation. A LIBS system equipped with a self-developed micro nanosecond laser was used to analyze the fragments of ceramic, and the principal components and loadings of ceramic spectral data were extracted using principal component analysis (PCA) to analyze the ancient ceramic fragment samples. The analysis results show that the ceramic fragments can be identified by using PCA, but similar element compositions cannot be accurately distinguished. Further classification using linear discriminant analysis (LDA) show that the spectra of the ancient ceramic fragment samples can be correctly classified. Finally, LDA classification algorithm was used to classify and identify all 18 ceramic debris samples, and the total classification accuracy is 98.15%. The results show that the use of micro-LIBS equipment can quickly and accurately identify cultural relics fragments, which can provide effective help for the protection and restoration of cultural relics fragments.