With the rapid development of 3D printing technology, fused deposition modeling is the most widely used additive manufacturing process. Due to its forming principle and device structure, humps, cracks, holes and other defects are generated due to large accuracy errors in the printing process, which not only affects the appearance and performance of parts, but also causes waste of time and materials. In this paper, a method for detecting side surface defects of FDM 3D printing products is proposed, which integrates machine vision, image recognition and other related technologies. Aiming at the problem that cracks and holes that may be encountered in the printing process may cause the instability of the performance of the workpiece, the identification, judgment and early warning are carried out. The follow-up printing work of the defective workpiece is suspended to avoid the waste of time, greatly liberate manpower and improve the yield of 3D printing.