摘要
为提高变电站巡检机器人进行指针式仪表自动读数的准确性和实时性,提出一种基于OpenVINO的变电站指针式仪表读数方法。通过关键算法的改进及OpenVINO工具的加速,样机系统对同质量图片的检测和识别速度达到78 ms,相对于TensorFlow框架提升了约40倍。通过对比改进前后Mask R-CNN的表盘分割效果、表盘矫正前后的表盘分割效果对仪表读数精确度的影响,可知在仪表自动读数阶段,仪表读数的准确度得到有效提升。
In order to improve the accuracy and real-time performance of automatic instrument reading by substation inspection robots,an OpenVINO-based method for reading pointer meters in substations is proposed.Through the improvement of key algorithms and the acceleration of OpenVINO tools,the prototype system achieves 78 ms detection and recognition speed for the same quality images,which is about 40 times better relative to the TensorFlow framework.By comparing the effect of improving the dial splitting effect of Mask R-CNN before and after,and the influence of the dial splitting effect before and after dial correction on the accuracy of instrument reading,it can be seen that the accuracy of instrument reading is effectively improved in the automatic reading stage of the instrument.
作者
陆文健
刘三军
来国红
LU Wenjian;LIU Sanjun;LAI Guohong(College of intelligent science and Engineering,Minzu University,Enshi 445000,China;College of Physical Science and Technology,Central China Normal University,Wuhan 430070,China)
出处
《电工技术》
2023年第5期20-26,共7页
Electric Engineering
基金
国家自然科学基金“CCFD非线性自干扰模型及高抑制比技术研究”(编号61961016)
湖北省自然科学基金“5G同频同时全双工(CCFD)自干扰消除模型研究”(编号2019CFB593)
湖北民族大学博士科研启动基金项目“针对OFDM数字基带发生器的IFFT处理器”(编号MY2018B08)。
