摘要
在无人机电网巡线的研究中,为提高绝缘子识别的正确率,结合航拍图像和绝缘子串的特点,提出了一种粗细定位相结合的航拍绝缘子图像分步识别方法。利用灰度化和图像增强对绝缘子图像进行预处理,基于形态学算法和连通域思想改进传统最大类间方差(OTSU)阈值分割算法并分割图像背景;提取绝缘子图像骨架,针对绝缘子串骨架为一条近似直线的特点及与其他输电设备(如杆塔等)骨架的差异,通过检测直线进行绝缘子串粗定位;计算粗定位区域不变矩特征,运用由Ada Boost算法训练的分类器进行对粗定位区域遍历识别,实现绝缘子串细定位。实验结果表明,基于骨架提取的绝缘子分步识别方法能够有效地识别出航拍图像中的绝缘子串,与单独利用图像特征和分类算法直接识别绝缘子的方法相比,其表现得更稳定,正确率有显著提高,更适用于无人机电网巡线。
A new two-stage recognition method of aerial insulators according to the characteristics of aerial images and insulator strings in transmission line patrol of unmanned aerial vehicles （UA-Vs） is proposed in this paper. This method is applied to improve the correct recognition rate of the insulator. First, gray processing and image enhancement were used to preprocess. The OTSU threshold segmentation algorithm was improved to segment the aerial images based on the mathematical morphology algorithm and connected domain algorithm. Then, the skeleton of images was extracted. According to the characteristic that insulator string skeleton is a approximate linear and the skeleton characteristics of other transmission facility such as towers etc, detecting the straight line of images was used to realize rough position of insulators. Invariant moment features of the candidate region were calculated and then insulators were recognized accurately by AdaBoost classifiers. The test results show that this method can effectively recognize insulators from aerial images. Compared with direct recognition using image features and classification algorithm solely, the proposed two- stage recognition method based on skeleton extraction exhibits more stable performance and obtains higher recognition accuracy, and is thus more suitable for transmission line inspection.
出处
《华北电力大学学报（自然科学版）》
CAS
北大核心
2015年第3期105-110,共6页
Journal of North China Electric Power University：Natural Science Edition
基金
中央高校基本科研业务费专项资金资助项目(2014MS140)
