Wheat is prone to insect infestations during harvesting, transportation, and storage, leading to heat, mold, and deterioration. Timely pest detection is vital for effective prevention and improved storage quality. Traditional methods, such as manual identification and biological information detection, have limitations, including low efficiency, grain damage, and difficulty in identifying pest larvae. This study proposed a method for detecting Sitophilus zeamais (S. zeamais) in the interior of wheat based on computed tomography technology and the Multi-feature and Vision Transformer U-Net model. The U-Net was enhanced with the Multi-Feature Extraction block and the Residual Vision Transformer block. After 200 training iterations, the model achieved a mean Intersection over Union of 94.4%. To use image processing technology to segment S. zeamais, create 3D models, and extract features such as volume, surface area, and length. S. zeamais develops through stages: egg, early larva, late larva, pupal, and adult. From epidermal erosion into the endosperm, it transitions from a round egg stage to a columnar shape and then develops various organs. The volume of the S. zeamais increases from 0.008 to 0.018 mm³ during the egg stage to 0.89 to 1.16 mm³ in the adult stage, and its length grows from 0.176 to 0.284 mm during the egg stage to 2.416 to 2.865 mm in the adult stage.This method offers accurate, rapid extraction and visualization of S. zeamais developmental information, supporting early-stage variation analysis and enhancing wheat quality and pest control.