基于多算法融合的破损陶瓷绝缘子图像识别与检测方法

Identification and Detection of Damaged Ceramic Insulators Based on Multi-algorithm Fusion

为了提高输电线路安全稳定运行,在传统检测绝缘子缺陷的基础上,提出一种多算法融合的破损陶瓷绝缘子检测方法。基于FASTER-RCNN算法,将拍摄到的绝缘子图像进行训练,得到绝缘子串具体位置,并改进SBGFRLS(线性自适应滤波)算法,实现对瓷绝缘子破损的精确识别。首先利用FASTER-RCNN算法,选择Faster R-CNN网格作为检测器,识别出利用无人机拍摄得到的图像中的绝缘子串所在位置;然后,利用SVM(支持向量机)算法对绝缘子图像进行粗分割,得到粗分割后的绝缘子缺陷识别图,再在粗分割的基础上,运用改进的SBGFRLS算法对图像中的绝缘子进行精确识别,得到破损绝缘子定位图。实验结果表明,得到的绝缘子破损识别图可以验证该改进算法的可行性和有效性。

In order to improve the safe and stable operation of transmission lines,a multi-algorithm fusion detection method for damaged ce-ramic insulators is proposed based on the traditional detection of insulator defects.Based on FASTER-RCNN algorithm,the cap-tured insulator image is trained to obtain the specific location of insulator string,and the SBGFRLS algorithm is improved to achieve the accurate identification of ceramic insulator damage.Firstly,FASTER-RCNN algorithm is used to select Faster R-CNN grid as a detector to identify the location of insulator strings in images captured by UAV.Then,the SVM algorithm is used to roughly segment the insulator image,and the insulator defect recognition map after rough segmentation is obtained.On the basis of rough segmentation,the improved SBGFRLS algorithm is used to accurately identify the insulator in the image,and the damaged insulator location map is obtained.The experimental results show that the obtained insulator damage identification diagram can verify the feasibility and effectiveness of the improved algorithm.