Circular objects detection in digital images, a vital and recurring problem in image processing and computer vision, has many applications especially aiding vectorization of line drawing images, pupil and iris detecti...Circular objects detection in digital images, a vital and recurring problem in image processing and computer vision, has many applications especially aiding vectorization of line drawing images, pupil and iris detection, circular traffic sign detection, and so on. In this paper, we lever Midpoint Circle Algorithm to speed up circle validation with sub-pixel precision. At the same time, we combine the least square method to improve the accuracy of circle detection. Experimental results from tests on synthetic and natural images validate that the proposed technique is efficiency regarding accuracy, speed and robustness.展开更多
针对配电网中的谐波与间谐波联合检测问题,提出一种基于深度学习的配电网谐波与间谐波联合检测方法,结合卷积神经网络(Convolutional Neural Network,CNN)与长短期记忆(Long Short Term Memory,LSTM)网络构建混合模型,实现谐波和间谐波...针对配电网中的谐波与间谐波联合检测问题,提出一种基于深度学习的配电网谐波与间谐波联合检测方法,结合卷积神经网络(Convolutional Neural Network,CNN)与长短期记忆(Long Short Term Memory,LSTM)网络构建混合模型,实现谐波和间谐波的高效联合检测。该方法通过提取时频特征,提高了谐波与间谐波的检测水平,并减少了计算负担。实验结果表明,该方法在检测精度和实时性方面均优于传统检测方法,为配电网的电能质量监测提供了有效的技术支持。展开更多
文摘Circular objects detection in digital images, a vital and recurring problem in image processing and computer vision, has many applications especially aiding vectorization of line drawing images, pupil and iris detection, circular traffic sign detection, and so on. In this paper, we lever Midpoint Circle Algorithm to speed up circle validation with sub-pixel precision. At the same time, we combine the least square method to improve the accuracy of circle detection. Experimental results from tests on synthetic and natural images validate that the proposed technique is efficiency regarding accuracy, speed and robustness.
文摘针对配电网中的谐波与间谐波联合检测问题,提出一种基于深度学习的配电网谐波与间谐波联合检测方法,结合卷积神经网络(Convolutional Neural Network,CNN)与长短期记忆(Long Short Term Memory,LSTM)网络构建混合模型,实现谐波和间谐波的高效联合检测。该方法通过提取时频特征,提高了谐波与间谐波的检测水平,并减少了计算负担。实验结果表明,该方法在检测精度和实时性方面均优于传统检测方法,为配电网的电能质量监测提供了有效的技术支持。
