| 利用卷积神经网络进行绝缘子自动定位 |
| |
| 引用本文： | 彭向阳,刘洋,王柯,张泊宇,钱金菊,陈驰,杨必胜.利用卷积神经网络进行绝缘子自动定位[J].武汉大学学报(信息科学版),2019,44(4):563-569. |
| |
| 作者姓名： | 彭向阳 刘洋 王柯 张泊宇 钱金菊 陈驰 杨必胜 |
| |
| 作者单位： | 1.广东电网公司电力科学研究院, 广东 广州, 510080 |
| |
| 基金项目： | 南方电网公司重大科技专项GD-KJXM20150903国家自然科学基金41531177中国博士后科学基金2016M600614 |
| |
| 摘 要： | 提出一种基于二值化赋范梯度特征和卷积神经网络的航空影像绝缘子自动定位方法。首先利用二值化赋范梯度分类器提取绝缘子候选窗口，而后利用卷积神经网络算法进行精细识别，获得包含绝缘子目标的窗口集，最后对高重叠度窗口集进行加权迭代合并得到最终绝缘子定位结果。采用广东电网大型无人机实际线路巡检获取的可见光影像对自动定位算法进行验证，实验结果表明，在复杂背景下绝缘子整体回调率为90.5%，定位精度为92%，证明该方法能够对复杂背景下可见光影像中的绝缘子进行有效识别定位，且算法通用性较强，可适应不同背景的可见光影像。
|
| 关 键 词： | 二值化赋范梯度 卷积神经网络 加权迭代合并 绝缘子 自动定位 |
| 收稿时间： | 2017-06-13 |
An Automatically Locating Method for Insulator Object Based on CNNs |
| |
| Affiliation: | 1.Guangdong Electric Power Research Institute, Guangzhou 510080, China2.State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079, China |
| |
| Abstract: | In this paper, a method is proposed to locate the insulator automatically in aerial image based on binarized normed gradients (BING) and convolutional neural networks (CNNs). Firstly, we extract insulator candidate windows with BING algorithm. Secondly, we identify the windows containing insulator with convolution neural networks. Finally, the weighted iteration of the window set with high overlap is used to acquire the final insulator positioning results. The method proposed in this paper is validated with the transmission line aerial images obtained by the actual inspection of the large-scale unmanned helicopter of Guangdong Grid Co. Experiment shows that the recall of insulators with complex background is 90.5% and the positioning accuracy is 92%, which means the proposed method can effectively locate the insulators in aerial image with complex background, the method also has strong versatility, and can be adapted to the visible light image of different background. |
| |
| Keywords: | |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《武汉大学学报(信息科学版)》浏览原始摘要信息 |
|
点击此处可从《武汉大学学报(信息科学版)》下载全文 |
|
