The acquisition of such agricultural information as crop growth and output is of great significance for the development of modern agriculture. Using the image analysis is important to gain information on plant properties, health and phenotype. This study uses the unmanned aerial vehicle images about Maize breeding material collected in Beijing Xiao Tang mountain town in June 2017. The four color space transformation of RGB, HSV, YCbCr and L*A*B was used to divide the UAV image foreground (crop) with the background (soil background), and the classification of two values was obtained. The morphology of maize seedling was identified by skeleton extraction, and the morphological structure was refined by adding noise removal. According to the growth of crops, crops are divided into two categories (multiple leaves, few leaves).The results show that the Harris corner detection method has the highest recognition accuracy, the recognition rate of less leaf type reaches 96.3%, the recognition rate of multi leaf type reaches 99% and the overall recognition rate is 97.8%. The accuracy of the traditional image recognition is improved by 2~3%, and the accuracy is reliable.When dealing with multi leaf individual plants, the accuracy of identification can reach 99%. At the same time, under the overlapping overlapping of multiple leaves, the research method in this paper has good applicability.