Mineral resources play an important role in the development of industry and national economy. However, with the expansion of mining scale, more and more abandoned mining land is formed due to resource depletion and poor management. Due to the prolonged mining impact, a large amount of heavy metal elements are present in the soils of mining wastelands. In such contaminated areas, high levels of heavy metals may have an impact on the environment and the human body. Land reclamation is an important method for remediation of contaminated and degraded soils. The detection of heavy metal content in the reconstructed soils is an important indicator of land reclamation efficiency and requires long-term follow-up and monitoring. The traditional chemical detection methods are inefficient and costly, and can not detect a wide range of heavy metals. Hyperspectral technology is a new technology with great potential for development and has a wide range of applications in environmental protection, resource utilization and regional sustainable development. After the rapid development in recent decades, the accuracy of instruments has been gradually increased, and the detection methods have gradually become mature, so as to realize the high efficiency of soil heavy metals. Easy detection provides a new way. Normal soil heavy metal content is generally relatively lower, and the use of spectral techniques to measure heavy metal content is more difficult, but mining iron ore mining area due to the soil more iron, will make the soil heavy metals in the form of existence and aggregation changes, impact the response of heavy metals to the spectra, and make the correlation between soil spectral reflectance and heavy metal content even more pronounced. The contents of heavy metal (As, Cr, Zn) in soils were obtained by sampling chemical detection method in the study area of reclamation mining area in Daye City, Hubei Province. The soil reflectance was obtained by means of FieldSpec4 spectrophotometer (350～2 500 nm) First-order differential, reciprocal logarithm, and continuous unmixing method were used to preprocess the reflectance curve respectively, and the spectral characteristic bands were extracted. The correlations between the three heavy metal elements and spectral features were analyzed and a stepwise regression model was established. The results showed that compared with the general soil, spectral data preprocessing could make spectral characteristic bands more obvious, of which the first-order differential and continuous removal were the most obvious. The characteristic bands of the three heavy metal elements were 495, 545, 675, 995, 1 425, 1 505, 1 935, 2 165, 2 205, 2 275 and 2 355 nm. Correlation analysis between soil heavy metal content and spectral characteristic bands showed that all the three heavy metals showed correlation with spectral curve, and most of the correlation coefficients reached above 0.5 and the maximum correlation coefficient was 0.663, and different heavy metal species and treatment methods led to significant differences in the correlation coefficients. Three heavy metal inversion models were established based on the characteristic bands with the highest correlation with heavy metals in soil. The optimal inversion model for each heavy metal was selected based on the size of inversion model r. Because of different selection of heavy metal species, Cr, Zn First-order differential step-by-step regression was the best inversion model, and heavy metal As continuous removal method gradually regression was the best inversion model. Through the test, Cr in the three kinds of heavy metals was the best, and RMSE is 2.67, followed by Zn, and As is the worst. Comparing the current different detection methods, we can see that hyperspectral inversion of soil heavy metal content spectrometer based on soil samples and spectral data pretreatment is ideal. The related research results can provide reference for the hyperspectral inversion of heavy metals in mining-abandoned soils.