Abstract: Vision based powerline inspection robot navigation plays an essential role in modern smart grid, power system, and powerline inspection. This paper presents a fast and accurate online object classification approach using the combination of Histogram of Oriented Gradient（HOG） features and Support Vector Machine（SVM）. Compared with traditional environmental constricted geometric based methods, the proposed approach significantly improves the classification performance without losing computing efficiency. In particular, the method works satisfactorily in large-scale complex background and illumination variation scenario. Experiment results show that the proposed approach can achieve the balance between accuracy and efficiency.

Key words: transmission line inspection, object recognition, directional gradient histogram, support vector machine

摘要： 基于视觉导航的输电线路巡检机器人在智能电网和输电线路巡检中有广泛的应用。提出一种基于主元方向梯度直方图特征的快速分类检测方法用于在线障碍物检测与识别。与传统基于几何结构基元的方法相比，该方法能在不提升计算复杂度的情况下显著提升识别精度，明显改善了识别算法在野外大范围复杂背景与光照影响下的性能。实验结果表明，该方法能达到精度和速度的性能平衡。

关键词: 输电线路检测, 目标识别, 方向梯度直方图, 支持向量机