Taking the SPOT5 satellite images of Heilongjiang Province and national forest inventory data as data source, the image texture, spectral features and topographic information of the sample plots were obtained, forest information of sample plots survey data was extracted as the true value. Through multiple linear regression analysis, forest structural parameters (forest stand quadratic mean diameter, basal area, stand volume and species diversity index) estimation models were established by combination of texture, spectral features and topographic information as independent variables, so as to select the optimal texture feature generation window and the optimal forest structure parameter inversion model. The results indicated that the texture, spectral features of SPOT5 images and forest structure parameters had a strong correlation, and the 9×9 window was the optimal texture generation window;the accuracy of the model was greatly improved after introducing the topographic factor. The estimation model of tree species diversity index R2adj was more than 0.72, and stand volume estimation model was the optimal model with R2adj of 0.864 and root mean square error of 21.260m3/hm2. The study suggested that using high resolution satellite image texture, spectral and topographic features to estimate the forest structural parameters had good application prospect.