Accurate, rapid and non- destructive estimation of apple canopy SPAD (soil and plant analyzer development) values using hyperspectral remote sensing has important significance for monitoring the growth of apple trees. Spectral reflectance and SPAD values of 196 apple canopies are measured in 62 orchards in two years. The original spectra and the continuum- removed spectra are analyzed and correlated with SPAD values. The continuum-removed spectral indices of DVI (difference vegetation index), RVI (ratio vegetation index) and NDVI (normalized difference vegetation index) are calculated by the combination of arbitrary two bands in the range from 350 nm to 1300 nm. The stepwise regression and principal component analysis methods are used to extract the principal components as the independent variables. A support vector machine regression model for estimation of SPAD values is constructed and verified by the data collected in the second year. The vegetation indices of the biggest correlation between SPAD and the three newly built vegetation indices are NDVI (406, 563), RVI (406,565) and DVI (646,695), and the correlation coefficient (r) reaches 0.677, 0.690 and 0.711, respectively. The support vector machine regression model is verified. Relative error of prediction (REP) is 1.190%, PR2 is 0.837 and relative percent deviation of prediction (RPDP) is 2.213. The continuum-removed method and building vegetation indices can improve the correlation between spectra and apple canopy SPAD values. The spectral index which has the highest correlation with SPAD is composed of the visible bands. The support vector machine regression model has excellent ability to estimate the SPAD values of apple canopies which are growing rapidly.