Abstract: In order to sex the pupae of two main tobacco pests for effective control, Helicoverpa assulta and Helicoverpa armigera, pupae images of Helicoverpa assulta and Helicoverpa armigera were analyzed based on machine vision processing with pattern recognition technology. An SLR camera was used to photograph the pupae of the two pests and the effective regions of abdominal end segments were extracted. The original images with a resolution of 350×350 pixel were obtained, and the gray images of the R channel in RGB space were used as the input for texture features. The extracted texture features such as contrast and angular second moment based on gray level co-occurrence matrix were taken as the basis of pupa sexing. The data of pupa features were sent to the trained support vector machine for sexing, and the results showed that this method could effectively sex the pupae of Helicoverpa assulta and Helicoverpa armigera with the recognition rates of 87.5% and 82.5% respectively. This method provides a technical means for machine recognition of pest pupae.