A study was conducted to address the challenges of inconvenience in detecting and controlling stress during the assembly of critical metal components in aircraft assemblies. The study focused on in-situ assembly stress detection technology based on image recognition. Employing a non-contact binocular stereo vision measurement approach, the research accurately characterized the surface micro-strain during the assembly of metal parts. Subsequently, the assembly stress of the metal components was derived through precise calculations. Moreover, a vision-based non-contact full-field strain measurement system was developed to detect the in-situ assembly stress of key components. This system ensured a precision of global strain detection within 250 micro-strain, with a measurement range extending from 0.025% to 200%. Additionally, a vision-based non-contact in-situ measurement approach was proposed for controlling stress in critical assembly areas. The application of the developed detection system in aircraft fuselage crack control has been successful, providing significant support for achieving low-stress and high-performance assembly of critical components and assemblies. This research contributes valuable insights into enhancing the precision and reliability of aircraft assembly processes.