Addressing the current limitations of traditional industrial cameras, which cannot achieve high frame rate dynamic capture and deformation measurement at high resolutions due to hardware cost constraints, this study investigates and designs a lightweight video frame interpolation neural network architecture based on optical flow motion prediction; A hybrid attention mechanism was employed to extract motion details, reducing the computational complexity of grayscale images and enabling image interpolation for low-cost industrial cameras. The results show that the percentage error of pixel point continuous frame motion after image frame interpolation does not exceed 5%, and the dynamic deformation interpolation error at high resolution is at the sub-millimetre level. The lightweight grayscale image frame interpolation network proposed in this paper provides a low-cost technical solution for dynamic deformation acquisition in industrial vision.