Optimization of feature space and improvement of the segmentation algorithm are the keys of accurately obtaining seedling information using object-oriented technology. An improved edge segmentation algorithm was used to segment image based on dealing with the noise of multispectral images. The algorithm developed the simulated balloon expansion method, and could control the direction of the force field, so that the curves were made to split and collapse inwards. And the feature space made up of texture, shape, spectral features was built to accomplish seedling information extraction. The results showed that the total accuracy of seedling information extraction was 86% by the method of this paper, 12% higher than that of traditional methods, and the KAPPA coefficient was 0.8145, 0.1159 higher than that of traditional methods. The method of this paper could accomplish seedling information extraction quickly and accurately, and provide a reference for the accurately monitoring and decision making to management departmen. It has important meaning to forecast and evaluation for the future afforestation situation.