Abstract: Aim at the welding defects, such as instability, due to parameter mismatch and assembly error during welding process, an online system for monitoring quality of laser welding was built based on coaxial image sensing technology to collect, which analyzed and extracted the characterisitic information from the welding pool during laser welding process. Results showed that instability, collapse defect and weld misalignment were associated with the characteristic of the welding pool in the condition of laser welding unequal thickness stainless steel sheet with laser power of 1 500 W. Combining BP neural networks algorithm, the monitoring system can realize the automatic identification and alarm function of the corresponding defects based on LabVIEW platform.