With the deepening of industrialization and urbanization, urban soil heavy metal pollution is becoming more and more serious. At the same time, traditional laboratory chemical analysis methods such as inductively coupled plasma mass spectroscopy have long analysis cycles and are prone to secondary pollution of the environment by experimental waste reagents. Portable X-ray fluorescence spectrometry is a testing method that can be used for rapid and non-destructive analysis directly in the field, and matrix effect is the most important factor affecting the testing accuracy and precision. The more commonly used calibration method is the traditional linear regression method, which is influenced by outlying values and still has large deviations in the processed data. The study attenuated the matrix effects during testing by adding the data of the major elements to the correction equation of the elements to be tested. In this study, heavy metals of Cr, Ni, Cu, Zn and Pb in soil samples from each campus of Jilin University was rapidly tested by portable X-ray fluorescence spectrometry under in situ to investigate the major elements that had the greatest influence on the matrix effect of each heavy metal element. The original Sherman equation was adjusted by combining the partial least-square method and the multiple linear regression method, and the new equation was used to correct for the matrix effect of each heavy metal element under the data of inductively coupled plasma mass spectrometry a reference. The differences between the data processed by this method and the traditional linear regression method were compared by statistical parameters, and the correlation between elements and samples was also analyzed by correspondence analysis. The results show that the major elements are the important factor affected by the matrix effect, and the matrix effect correction equation based on different major elements is effective, with applicability Cr>Pb>Zn>Ni>Cu. The quality of the corrected data was significantly improved, the coefficient of determination increased, the regression images were concentrated, the mean absolute error and root mean square error was further reduced. The correction effect was better than the traditional linear regression method. The matrix effect correction method mainly reduces the overall average error and discrete degree of the data by reducing the deviation of outlying values. The processed data meet the quantitative analysis requirements and can be extended to portable X-ray fluorescence spectrometry for rapid large area testing of heavy metals to detect environmental quality. At the same time, correspondence analysis is an analysis method between multi-dimensional data dimensions and multi-dimensional data dimensions. It has excellent results for classification and correlation analysis between multiple variables.