CAJ | 학술논문

为了提高铁路扣件检测的识别率和鲁棒性，以及扣件图像PHOG特征的有效性，提出了简单有效的枕肩定位算法，该算法首先在提取PHOG特征前，根据枕肩、扣件和背景间的位置关系去除冗余背景信息；然后，模拟人眼视觉注意机制，设计MSLBP特征采样方式，提取扣件图像的宏观纹理特征；最后，采用分层次加权融合的方法联立两类特征，并采用SVM分类器进行扣件分类识别，提出一种基于计算机视觉和PHOG-MSLBP融合特征的缺陷识别算法．将该算法应用于实验，结果表明：与使用PHOG、MSLBP单一特征相比，基于PHOG-MSLBP融合特征检测算法的平均识别率分别提高了6．3％、4．5％，且鲁棒性更强，可满足扣件缺陷自动化检测的需要．
위료제고철로구건검측적식별솔화로봉성，이급구건도상PHOG특정적유효성，제출료간단유효적침견정위산법，해산법수선재제취PHOG특정전，근거침견、구건화배경간적위치관계거제용여배경신식；연후，모의인안시각주의궤제，설계MSLBP특정채양방식，제취구건도상적굉관문리특정；최후，채용분층차가권융합적방법련립량류특정，병채용SVM분류기진행구건분류식별，제출일충기우계산궤시각화PHOG-MSLBP융합특정적결함식별산법．장해산법응용우실험，결과표명：여사용PHOG、MSLBP단일특정상비，기우PHOG-MSLBP융합특정검측산법적평균식별솔분별제고료6．3％、4．5％，차로봉성경강，가만족구건결함자동화검측적수요．
In order to improve the recognition rate and robustness of railway fasteners detection,and to increase the effectiveness of PHOG (pyramid histogram of oriented gradients)feature,a simple and effective sleeper shoulder locating algorithm was proposed. In this algorithm,the redundant information in fastener images was removed before extraction of PHOG feature, according to the positional relationships among the sleeper shoulder,fasteners,and background. Then,an MSLBP (macroscopic local binary pattern )sampling method was designed and applied to extract the macroscopic texture feature of the fastener images,which could well simulate human visual attention mechanism. Finally, the two different categories of features were integrated by the hierarchical weighted fusion method;using the SVM classifier to classify and detect fastener defects,a defect recognition algorithm based on computer vision and PHOG-MSLBP fusion feature was presented. The algorithm was applied to experiments,and the results show that the average recognition rate based on PHOG-MSLBP feature is 6. 3% higher than that based on PHOG feature,and 4. 5% higher than that based on MSLBP feature. In addition,the proposed algorithm is more robust than several mainstream methods,and can meet the need of automatic inspection of fastener defects.