The internal components and circuits of power supplies are diverse and complex. When detecting connections, they are easily affected by background interference, which affects visual judgment. On the other hand, wires are deformable materials and can be arranged and stored in different ways depending on the assembler. These factors make it difficult for both manual and traditional optical inspection to be performed, making it difficult to effectively control product quality on the production line.Using Solomon SolVision's Segmentation technology, the correct and incorrect feature patterns are defined according to the wire color and terminal block assembly conditions in the image, and the AI model is trained. The trained AI model can accurately detect and locate wire misconnection defects and identify defective products in real time.