摘要
提出了一种电网巡检航拍图像中绝缘子定位方法。该方法利用遗传算法改进的最大类间方差法对航拍绝缘子图像进行阈值分割,然后基于形态学滤波重构;采用Sketch Up软件构建绝缘子图像库,即正样本,对该正样本和负样本(背景)分别提取Hu不变矩来训练Ada Boost分类器;利用训练好的分类器对分割后的绝缘子进行分类和粗定位;通过顺序割图法(Order Cut method,OC)对粗定位得到的包含同一个绝缘子的所有矩形的并集的初始最小外接矩形进行裁割,最终获得精细定位的完整绝缘子。实验结果表明,该方法能从航拍图像中有效、精确地定位不同类型、不同长度的完整绝缘子。
A method to locate the insulators in the power grid patrol aerial images was proposed. OTSU algorithm improved by genetic algorithm was applied to segment insulators from backgrounds. The segmented images were filtered and reconstructed based on morphology algorithm. SketchUp software was used to build insulator samples, namely positive samples. The moment invariants of these samples and negative samples（background） were calculated respectively to train a classifier based on AdaBoost algorithm. By using the trained classifier, segmented insulators were classified and located coarsely. The Order Cut method （OC） was used to cut the original minimum circumscribed rectangle consisting of all rectangles which located the same insulator, and the insulators could be located exactly and completely. The testing results show that this methodology can locate different types and length of insulators from aerial images effectively, accurately and completely.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2017年第6期1327-1336,共10页
Journal of System Simulation
基金
国家自然科学基金(61472419)
中央高校基本科研业务费专项基金(2014MS140)
关键词
航拍图像
绝缘子定位
阈值分割
不变矩
顺序割图法
aerial images
insulators location
threshold segmentation
moment invariants
Order Cut method
