当前位置: X-MOL 学术Electr. Eng. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
An intelligent recognition system for insulator string defects based on dimension correction and optimized faster R-CNN
Electrical Engineering ( IF 1.6 ) Pub Date : 2020-09-09 , DOI: 10.1007/s00202-020-01099-z
Tao Lin , Xiaowei Liu

In this paper, an intelligent recognition system for insulator based on dimension correction and optimized faster region with convolutional neural network (R-CNN) is proposed. In the process of insulator pictures shooting, a laser radar is used to calculate the UAV correction vector. The position of the UAV is adjusted to ensure the consistency of the spatial dimensions of the pictures taken in different time dimensions. Based on the almost invariant spatial dimension, the faster R-CNN image recognition algorithm is optimized. When the target detection frame is generated, marked reference pictures are added to narrow the search range, improve the target detection frame generation speed, and reduce the number of pictures during training. Experiments and comparison analysis are included. They verify the optimized faster R-CNN image recognition algorithm requires less pictures and recognition time, and the recognition accuracy increased from 85.6 to 97.3%.

中文翻译:

基于维数修正和优化faster R-CNN的绝缘子串缺陷智能识别系统

本文提出了一种基于尺寸校正和卷积神经网络优化快速区​​域的绝缘子智能识别系统(R-CNN)。在绝缘体图片拍摄过程中,利用激光雷达计算无人机修正矢量。调整无人机的位置,保证不同时间维度拍摄的图片空间维度的一致性。基于几乎不变的空间维度,优化了更快的R-CNN图像识别算法。在生成目标检测框时,添加标记的参考图片,缩小搜索范围,提高目标检测框生成速度,减少训练时的图片数量。包括实验和比较分析。
更新日期:2020-09-09
down
wechat
bug