Detection method of high voltage bushing based on Hu moment invariants and SVM

In this paper, a new detection method of high voltage bushing is proposed using Hu invariant moments and Support Vector Machine (SVM), which is designed in regard to both complex environments of transformer substation and the characteristics of pictures obtained by inspection robots. The Niblack algorithm is first used as a binarization method to decrease the influence of changing of lighting. The morphology method is then used to remove messy interference while maintaining the morphological characters of the target, and a few potential regions of high voltage bushing can be found. Hu invariant moments is used to extract the features of the potential positions, and a SVM is trained to classify and recognize the positions of high voltage bushing. The experimental results show that the proposed detection method is effective and it can remove the complex background of the substation and reduces the influence of changed lighting, shooting position and shooting angle.