CAJ | 학술논문

针对荧光磁粉探伤中人工观察识别工作量大、效率低、难以保证检测质量一致性及健康危害严重等问题，提出一种具备机器自主学习能力的智能图像识别方法。结合人工识别磁痕缺陷的工艺方法分析磁痕图像，基于图像形态学提取荧光磁粉缺陷多项特征建立专家知识库，同时根据人员对图像识别结果的判定情况自动修正专家知识库。实验证明该方法可有效识别车削类工件中存在的缺陷，鲁棒性较强。
침대형광자분탐상중인공관찰식별공작량대、효솔저、난이보증검측질량일치성급건강위해엄중등문제，제출일충구비궤기자주학습능력적지능도상식별방법。결합인공식별자흔결함적공예방법분석자흔도상，기우도상형태학제취형광자분결함다항특정건립전가지식고，동시근거인원대도상식별결과적판정정황자동수정전가지식고。실험증명해방법가유효식별차삭류공건중존재적결함，로봉성교강。
An intelligent image recognition method with machine self - learning ability was proposed to o-vercome the shortcomings of heavy workload of artificial observation,low efficiency,unguaranteed consis-tency in detection quality and health hazards in fluorescent magnetic particle flaw detection. First,artifi-cial recognition magnetic image defect method was applied to analyze the magnetic mark image. Then the features of magnetic image defect were extracted based on image morphology to establish the expert knowl-edge base which is automatically corrected according to the personnel judgment of the results of image rec-ognition. The experimental results show that the proposed method can effectively recognize the defects of turning workpieces and has strong robustness.