Maize leaf traits have great significance to the study of growing development, breeding and functional gene research. However, the traditional method is inefficient, subjective, and also with less measurement, which is far from the requirement of maize-related research. Therefore, an automatic and dynamic technology for maize leaf traits extraction was proposed. Totally 100 maize varieties were adopted, and eight growth points were analyzed every three days based on the high-throughput crop phenotyping platform. For each measurement, the 18 side-view images were acquired every 10°, and the maximum side-view image was identified based on the width information. Then an improved segmentation method was applied to extract the complete plant binary image. After that a parallel thinning was used to extract the plant skeleton, and Hough transform was adopted to distinguish leaf skeleton from the stem. Finally, each leaf skeleton was labelled and the specific algorithm was developed to calculate the leaf length, angle and curvature. The experimental results showed that the measurement error for leaf length and leaf angle was 0.92% and 3.32%, respectively, and the results demonstrated that this method had a higher consistency than manual method. Since the new leaf would always grow from above in maize, the leaf matching based on time series was designed and carried out by using the leaf relative-position information. With leaf registration, the leaf growth rate and leaf curvature variation were obtained. In general, this study provided a novel method for maize-related research.