基于多层卷积网络的圆弧快速检测算法设计

Design of Fast Arc Detection Algorithm Based on Multi-layer Convolutional Network

针对当前圆弧检测算法存在误检率高、检测效率低的问题,提出一种基于多层卷积网络的圆弧快速检测算法。首先基于多层卷积网络进行圆弧信息的聚类存储,通过聚类方向自带的惯性作用将数目较少的孤立像素点当作噪声信息直接过滤,并再对其他圆弧进行全局聚类。然后以圆弧的圆心所在区域为依据对圆心阈值区域进行重构,获取圆弧阈值区间。最后根据优化策略对阈值区间进行优化,并对圆弧进行拟合和去伪,实现圆弧的快速检测。为了验证该圆弧快速检测算法的有效性,与基于切线段匹配的快速圆弧检测算法和基于边界聚类的圆弧检测算法进行对比,仿真实验结果证明该算法的检测效率更高,误检率远低于其余两种传统圆弧检测算法,并且噪声更小,清晰度更高。

Aiming at the problems of high false detection rate and low detection efficiency in current arc detection algorithms,we propose a fast arc detection algorithm based on multi-layer convolutional network.Firstly,based on multi-layer convolutional network,arc information is clustered and stored.Through the inertial action of clustering direction,isolated pixels with small number are directly filtered as noise information,and other arcs are globally clustered.Then the threshold area of the center of the arc is reconstructed according to the area of the center of the arc to obtain the threshold interval of the arc.Finally,the threshold interval is optimized,and the arc is fitted and falsified to realize the fast detection of arc.In order to verify the effectiveness of the fast arc detection algorithm,it is compared with the fast arc detection algorithm based on tangent segment matching and the arc detection algorithm based on boundary clustering.The simulation results show that the detection efficiency of this algorithm is higher,and the error detection rate is far lower than the other two traditional arc detection algorithms,with less noise and higher definition.