基于3D模型和AdaBoost算法的绝缘子检测

随着电力系统直升飞机巡线的不断发展与应用,对于输电线路关键部件的检测与识别越来越受到图像处理工作者的青睐。提出了一种利用3D模型制作训练样本及Ada Boost算法实现的航拍图像绝缘子自动检测方法。根据绝缘子3D模型图像的空间结构特征,提出了能反映这些结构的Haar矩形特征,从中挑选对绝缘子航拍图像有最好区分的特征构成弱分类器,再组合生成强分类器。使用正负样本图像训练后,由强分类器级联组成了一个多层分类器系统。实验结果表明,该方法有效地提升了绝缘子的识别效果,为后续的故障检测工作提供了良好的铺垫。

Insulator detection based on 3D models and AdaBoost algorithm

With the application and development of aerial surveillance by helicopters, the automatic extraction of insulators from aerial images gets more attention of photogrammetrists. A method for extracting insulators from aerial images is proposed in this paper by using 3D models creating training samples and AdaBoost algorithm According to structure features of 3D models in gray scale images, Haar features are presented and weak classifiers with easy distinguishing features are created, and then strong classifiers are created too. After training samples, strong classifiers are cascade connected to build a multiple layer classifier system . The test results show that this method improves the insulator detecting effectiveness of which makes a good basis of subsequent fault detection.