基于MM-Net的阀芯外表面缺陷检测

Defect detection on the outer surface of valve core based on MM-Net

阀芯是喷油器的重要组成部分,复杂的加工工艺与加工环境会导致部分阀芯出现多种形态的生产缺陷,传统的人工目检存在主观性强、检测效率低等缺陷。基于此,本文提出一种基于MM-Net的阀芯外表面缺陷检测方法,所提出的MM-Net检测模型由基于注意力机制的金字塔逐级多层次特征提取模块、多尺度特征融合模块等4个模块组成,多个模块的组合有效实现了对阀芯外表面缺陷特征的提取、提纯与融合,并在判别缺陷类型的同时实现了对瑕疵位置的定位。最后通过对比实验,验证了本文所提出方法在阀芯外表面缺陷检测方面的优势。

The valve core is an important part of the fuel injector. The complex processing technology and processing environment will results in various forms of production defects in some valve cores. The traditional manual visual inspection has defects such as strong subjectivity and low detection efficiency. Based on this, this paper proposes a MM-Net-based spool outer surface defect detection method. The proposed MM-Net detection model consists of four models including pyramid-based multi-level feature extraction modules and multi-scale feature fusion modules based on attention mechanism et al. The combination of multiple modules effectively realizes the extraction, purification and fusion of defect features on the outer surface of the valve core. While judging the defect type, the location of the defect position is realized. Finally, through comparative experiments, the advantages of the method proposed in this paper are verified in the detection of defects on the outer surface of the valve core.