采用非经典感受野交互的钢轨紧固件螺母中心多特征分级定位算法

A Multi-Feature Hierarchical Location Method for Railway Fastener Hexagon Nut Based on R-NCRF

针对铁路紧固件图像易受周边环境干扰、导致其几何中心定位精度不理想的问题,提出一种紧固件螺母中心多特征分级定位算法。该算法在优化蝶形非经典感受野模型的刺激区和抑制区的基础上,设计一种路径非感受野交互模型以匹配紧固件特征,具备优异的干扰噪声抑制能力,能精确提取紧固件圆轮廓和六角轮廓特征。根据轮廓特征,多特征分级定位算法利用六角边之间的角度约束和距离约束关系,实现了不完整轮廓下的中心精确定位。实验结果表明:该算法在正常条件下和受光照影响条件下使圆检测准确率分别提高了3%和24%,具备干扰环境下的紧固件中心精确定位能力。

A multi-feature hierarchical positioning algorithm for the fastener nut center is proposed to solve the problem that the railway fastener image is vulnerable to the surrounding environment and its geometric center positioning accuracy is not ideal. A route non-receptive field interaction model is designed to match the fastener features based on the optimization of the stimulation and suppression zones of the butterfly non-classical receptive field model. The proposed model has excellent interference noise suppression ability and can accurately extract the circular contour and hexagonal contour of a fastener. Based on the contour feature, the multi-feature hierarchical positioning algorithm utilizes the angular constraint and the distance constraint relationship among the hexagonal edges to achieve accurate center positioning from incomplete contour. Experimental results show that the accuracy of the circle detection under normal conditions and under the influence of illumination increases by 3% and 24%, respectively, and the algorithm is capable of locating the position of the fastener center accurately under interference environments.