Abstract: Scanning electron microscope（SEM） is an important instrument for critical dimensional measurement of semiconductor devices. In order to improve the measurement accuracy and the calibration efficiency of SEM, a set of line spacing samples with a nominal value of 100 nm to 10 μm and lattice samples with a nominal value of 2 μm to 10 μm was developed. In addition, a calibration method for SEM based on image processing was proposed, aiming at the length measurement and indication error, orthogonal distortion and linear distortion in the existing calibration specification. The experimental research shows that the developed line interval template and lattice samples have high quality parameters, which can cover the measurement range of SEM and solve the problem of instrument traceability. The calibration method for SEM based on image processing can effectively improve the measurement accuracy and the calibration efficiency.