基于Faster-RCNN及一维曲线分析的表计指针识读方法

Pointer Reading Based on Faster-RCNN and One-dimensional Curve Analysis

针对机器人部署调试中表计需逐个配置的问题,提出基于神经网络学习及一维曲线分析的表盘指针、刻度识别方法。基于Faster-RCNN(faster region-based convolutional network),初步识别刻度盘位置,从而确定指针运动圆心。将图像以圆心为中心进行极坐标展开后,通过分析一维曲线的频域特征筛选刻度所在区域,统计不同区域灰度峰谷值的形式确定刻度和指针的真实位置。最后对几种典型表计进行实验分析,验证所提方法的有效性。实验结果显示,该方法对表计样本量较多的表计类型,可成功实现自动读数功能。

According to the situation the dials need to be configured one by one in electrical robot deployment and debugging,an auto-reading method of dial pointer and scale is proposed based on neural network learning and one-dimensional curve analysis.Through learning and recognizing the dial position based on Faster-RCNN(faster region-based convolutional network),the motion center of the pointer is initially determined.After the image is expanded in polar coordinates with the center of the circle,the region where the scale is screened by frequency analyzing with the one-dimensional curves,and the true position of the scale and pointer is determined by using image gray statistical characteristic of different regions.Finally,the effectiveness of the proposed method is verified by experimental analysis of several typical dials.The results show the proposed method can successfully realize automatic reading function for large sample meter types.