Infrared photoelasticity method can achieve non-destructive and full-field characterization of internal stress in semiconductor materials. Most of the existing unwrapping method developed for infrared photoelasticity require some manual interventions. Lack of automated means without manual intervention is a key factor restricting the application of infrared photoelasticity to the on-line stress detection. This paper proposed an automatic full-field internal stress analysis technique based on dual-wavelength infrared photoelasticity, namely automatic seeds identification for dual wavelengths. According to the phase trend of each point in the sample at different wavelengths, the initial seeds required for phase unwrapping can be automatically identified by the symbol and absolute value of the real phase, and thus avoiding manual intervention in the whole process of photoelastic analysis. In this work, a dual-wavelength infrared photoelasticity device was developed based on the proposed technique. The results of application experiments were compared with those of other existing methods, which verified the reliability of the proposed method, and its device as well, for the automatic analysis of complex internal stress distribution.