The leachate generated from the long-term storage of waste residue and waste liquid produced in mining mineral resources diffuses into the soil, which is easy to cause the surrounding soil to be polluted by heavy metals and affects its crop growth. After human beings eat fruits containing heavy metals through the food chain, they will cause neurasthenia of the nervous system, numbness of hands and feet, indigestion of the digestive system, blood poisoning, kidney injury and other symptoms. Then it will pollute and damage the ecological environment and personal safety. therefore, how to quickly find out the situation of soil pollution is particularly critical. With the development of remote sensing technology, multispectral remote sensing has great potential in breaking through the vegetation barrier to monitor heavy soil metals because of its high spectral resolution and real-time non-destructive and large-area monitoring advantages. This study takes peach trees, the main crop in Pinggu District, as the research object. Using hyperspectral data of peach leaves and field soil sampling data, the response characteristics of peach leaf spectral curves were analyzed. The reflectance spectra of peach leaves were transformed by first-order/second-order differentiation, standard normal transformation and continuous de unification. The characteristic variables are determined by correlation analysis and multiple linear regression model, construction of vegetation index HMSVI, the correlation between HMSVI and Cd, As and Pb content is higher than that of common vegetation index. After modeling element content and vegetation index HMSVI by linear regression method, selecting the model with good fitting, the statistical modeling of leaf hyperspectral reflectance spectrum and soil heavy metal content was realized, and the spatial distribution of heavy metal content was retrieved from sentinel-2 remote sensing image, and the results were verified. The results show that: the average spectral reflectance of leaves under heavy metal stress was higher than that of normal leaves, and the phenomenon of “blue shift” occurred. 780, 945 and 1 375 are the most sensitive to heavy metal pollution. The inversion model established by using the vegetation index constructed in three bands can be better used to predict the content of heavy metal elements in peach forest soil. The prediction models are y=0.44x+0.193, y=7.436lnx+13.161, y=-15.359x+13.583x²+23.541 respectively. The spatial inversion results show that the high-value areas of the three heavy metals are widely distributed near the liujiadian tailings pond, Wanzhuang tailings pond and Jinhai Lake tailings pond in Pinggu District. Heavy metal pollution is more serious in the West than in the East. The mapping results can provide basic data support for preventing and treating heavy metal pollution in Taolin, Pinggu District, Beijing.