Heavy metal pollution constitutes one of the most urgent problems in soil environmental pollution, as plants become enriched in heavy metals through the soil, which endangers human health and poses a great potential danger to the ecological environment. The monitoring over heavy metal pollution in soil by traditional chemical methods is time-consuming and laborious and limited in scope. However, the method for monitoring heavy metal in soil leveraging hyperspectral vegetation technology is capable of quickly and accurately obtaining the heavy metal content in the soil, breaking through the vegetation barrier, and making the monitoring more efficient. Providing an important reference for the monitoring over and early warning of heavy metal elements in soil, this method matters for achieving the goal of constructing ecological civilization into a higher level and improving the quality of arable land. In this study, peach trees, the dominant economic fruit tree in Beijing, were research targets. 50 sampling points were evenly set up in the study area, and the spectral data of peach tree leaves were measured by using FieldSpec 4 portable ground wave spectrometers, while soil samples were collected and brought back to the laboratory for testing and analysis to obtain the data of heavy metal content in the soil. Efforts were made to analyse the leaf spectral characteristics of peach tree leaves under the stress of heavy metals in soil in different kinds of pollution and investigate how different soil heavy metals are correlated with leaf spectra through calculation. It was determined that element As in soil had a higher correlation with spectral reflectance. As a result, we calculated the correlation coefficients between element As in soil and vegetation indices, and construct a prediction model for elements As in soil using the appropriate vegetation indices. The results show that the spectral reflectance of peach leaves in the polluted area was generally higher than that in the background area and was more sensitive to heavy metals in soil in the wavelength range of 760~1 300 nm. The heavy metals in soil did not interfere considerably with the position of the red, blue and yellow edges of the leaves and were sensitive to the slope of the red, blue and yellow edges, and all of them were positively correlated. Spectral reflectance was weakly correlated with elements Cr, Cu and Hg in soil, and 0.1 level of significant correlation was reached with elements As, Pb and Cd in some wavelength ranges. The overall correlation curve trend was the same, with the correlation magnitude ranked as As>Pb>Cd in order. According to the above studies, it is found that As elements in soil have the strongest correlation. Therefore, we performed correlation analysis using As elements in soil and vegetation index, which showed that As elements were significantly correlated with both PRI1 and PRI3. The regression analysis was performed using SPSS data analysis software with PRI1 and PRI3 as independent variables and As in soil as a dependent variable. The test results show that the index prediction model of PRI3 (y=e^43.644x-39.386, R²=0.937, RMSE=0.161) rendered the best results and was more stable.