In this paper, a system is proposed for extracting insulators on power transmission lines from photographs captured by an unmanned aerial vehicle. The approximate regions of the insulators are first determined by a support vector machine with the histogram of oriented gradients as the feature descriptor. Then, the specific regions of insulators are detected and extracted by the GrabCut algorithm. In advance, some constraint conditions, such as the value ranges of color component values as well as the relationships between the component values in three color models, need to be specified. In our system, an interactive interface is developed to help determine these conditions. According to the experimental results, our system is capable of removing most of the backgrounds and extracting the insulators from photographs.

中文翻译：

---

使用支持向量机和颜色模型从无人机照片中提取输电线路绝缘子的视觉提取系统

在本文中，提出了一种系统，用于从无人机拍摄的照片中提取输电线路上的绝缘体。绝缘体的近似区域首先由支持向量机确定，以定向梯度的直方图作为特征描述符。然后，通过 GrabCut 算法检测并提取绝缘体的特定区域。需要预先指定一些约束条件，例如颜色分量值的取值范围以及三种颜色模型中分量值之间的关系。在我们的系统中，开发了一个交互式界面来帮助确定这些条件。根据实验结果，我们的系统能够去除大部分背景并从照片中提取绝缘体。