Heavy metal pollution in soil is becoming more and more serious. LIBS is one of the most promising technologies for rapid detection of heavy metal contamination in soil. However, due to the wide variety of soils and complex matrices, accurate quantification remains a challenge. In total, 451 soil samples were prepared and detected by the portable detector of LIBS, which were divided into six categories based on the compactness of the soil pellets, and a separate quantitative model for each type of soil sample was used for quantitative analysis by external standard method. It did not need a lot of data to train the model, and only a small number of calibration samples could be used for quantitative analysis of a large number of samples. The results showed that 78 standard samples and 334 collected samples were quantitatively analyzed by 39 standard samples. Compared with the standard value, the correlation coefficients were all above 0.95. A comparative experiment indicated that the portable LIBS system combined with soil classification and calibration methods can achieve fast and accurate quantitative detection.