Abstract: To forecast the outbreak of pest insects is based on the estimation of population density. To improve efficiency and accuracy of the current estimation system for cereal aphids, a novel method based on computer vision technology was introduced. The new method works in the follow ways: the first step was to set sticky yellow pans in wheat fields to trap aphids; secondly, digital images of the yellow pans were taken as the data source for the image analysis system. Then the images were processed for counting the numbers of aphids automatically using the image segmentation and connected region-labeling method. Experiments show that the yellow pan can trap cereal aphids effectively and can provide a plain background to facilitate the counting, so that the automatic counting system counts the tiny insects with accuracy of over 93.88%. This new method proves to be promising to substitute the current estimation system with old routine for collecting data of aphids from the field.