Abstract: To identify the defects of fusion welding joint, a magnetic sensing device was utilized to detect the weld block of S06 material based on the weak magnetic detection technology. A data processing algorithm was proposed to distinguish the stress concentration from the defect signal by using the threshold line of the magnetic induction gradient curve of the upper and lower magnetic coils. The results show that the magnetic induction curve of the magnetic coil has magnetic anomalies in both stress concentrations and defects while the magnetic induction intensity curve of the upper magnetic coil only demonstrate anomalies at stress concentrated sites. Additionally, at thosesites, the gradient curves of the upper and lower magnetic measuring coils both exceeded their respective threshold lines. Conversely, at defects, only the gradient curves of the lower magnetic measuring coils surpassed their threshold lines. Moreover, defect signals were analyzed by visualizing the two-dimensional image display, which could locate the defects more intuitively.