地基可见光全天空云图云量图像处理识别方法

Cloudiness Recognition Algorithm of Ground-Based Visible All-Sky Images Based on Image Processing

为增强地基可见光全天空云图中云与天空的特征和区别,提高云检测率,基于图像复原和图像增强技术提出一种改善云图质量的方法。该方法采用暗通道去雾算法进行图像复原;采用亮度直方图均衡增强图像纹理细节;综合两种方法,先图像复原,再图像增强。按低能见度薄云、低能见度厚云、高能见度薄云、高能见度厚云4种情况分别进行讨论,结果表明:除高能见度薄云采用单一的图像复原使云检测效果降低外,图像复原和图像增强都能使云检测和云量识别准确率提高;综合二者,云检测和云量识别准确率进一步提高;该方法对薄云和低能见度云图的改善最为显著。

In order to further distinguish the clouds and the background and improve cloud recognition,this paper proposes an algorithm to improve the quality of cloud images based on image restoration and enhancement techniques.First,a dark channel prior-dehazing algorithm is employed for image restoration.Then,the features are enhanced using the brightness histogram equalization algorithm.Finally,the two algorithms are combined by using image enhancement after image restoration.At the same time,the four conditions of thin cloud in low visibility,thick cloud in low visibility,thin cloud in high visibility,and thick cloud in high visibility are discussed,respectively.According to the simulation results,except that a single image restoration on thin clouds in high visibility has reduced cloud recognition,the proposed algorithm has significantly improved the image quality and cloud recognition.In addition,the image quality and cloud recognition can be further enhanced using the image restoration and enhancement techniques together,and there are more improvements on thin clouds and clouds in low visibility.