基于改进的Faster R-CNN高压线缆目标检测方法

Object detection of high-voltage cable based on improved Faster R-CNN

利用带电作业机器人取代人类的手动作业,可以有效地减少高电压、强电场对人体的危害,大大提高作业的效率。为解决带电作业机器人在复杂背景环境中对线缆目标的智能检测问题,提出基于改进的FasterR-CNN高压线缆目标检测方法。为了提高网络提取图像高级特征的能力,引入跳转连接并调整激活层、卷积层的顺序;然后对候选框生成机制进行改进,提升网络对小目标检测的性能;最后利用ROI池化层提取每个区域的特征,同时完成分类和框回归任务。通过构建高压线缆图像数据集,基于改进的Faster R-CNN模型进行大量实验,最后取得了较好的精度和较快的速度。

The use of live working robots to replace human manual operation can effectively reduce the harm of a highvoltage and strong electric field to the human body and considerably improve the working efficiency.To solve the intelligent high-voltage cable object detection problem for live working robots under a complicated background environment,a high-voltage cable object detection method based on the improved Faster R-CNN is proposed.To improve the capability of extracting the high-level features of images in the network,skip connections are introduced and the order of the activation and convolution layers is adjusted.Then,the proposal bounding box generation mechanism is improved to enhance the performance of the proposed method for small object detection.Finally,the features of each region are extracted using the ROI pooling layers,and the classification and bounding box regression tasks are accomplished at the same time.Through the construction of high-voltage cable image datasets and the performance of numerous experiments based on the improved Faster R-CNN model,good accuracy and fast speed have been achieved.