Classification And Recognition Of Ground-based Cloud Images Based On Multi-texture Features And Multi-scale Analysis

Cloud classification plays an important role in weather forecasting. Research on the status of cloud will help us to observe the process of atmospheric motion and predict weather conditions correctly. The successful implementation of cloud classification can improve the accuracy of weather forecasting. At present, the recognition of cloud still depends on visual observation, the results of recognition are inconsistent for some subjective factors and the different shape of cloud, this has become a bottleneck of the automation of meteorological service development. In recent years, with the development of image processing and pattern recognition theory, the study of automatic classification and identification of the cloud has become a hot topic of meteorological applications. The recognition of satellite cloud images and the recognition of ground-based cloud images are two main kinds of research in the area of cloud classification. It’s more suitable to describe the changes of a wide range of clouds by Satellite cloud images. Ground-based cloud images have great directive significance on the local cloud observations, unlike the former, the latter provides more texture information of the cloud, and it is easy to operate and low cost. So the automatic recognition of cloud based on ground-based cloud images has been a research hotspot.Feature extraction is the key problem in the areas of the recognition of ground-based cloud. It’s very important to extract effective features from ground-based cloud images to realize the recognition of ground-based cloud. In this paper, we did a more in-depth research and analysis of the texture feature of ground-based cloud, and proposed two improved methods and apply them to identify the cloud. Research content and main work in this paper are as follows:(1) The current research of ground-based cloud classification are analyzed, we found that there are a wide variety of cloud types and they are very complexity and diversity, it’s not a good texture features description of the cloud using single feature, resulting in lower recognition rate of ground-based cloud. This paper proposes a multi-texture features recognition method of ground-based cloud to solve this problem. Firstly, we extract the texture features of ground-based cloud by GLCM and Gabor wavelet algorithm, and then the extracted texture features are combined to get the final identification features. The final results show that the combination of features can describe the texture characteristics of cloud better and it’s also can improve the recognition rate of ground-based cloud.(2) The influence of light and rotation of the cloud lead to low robustness of features, and the traditional methods do not take the local characteristics of cloud into account. A multi-scale analysis identification method of the ground-based cloud is presented in this paper. Firstly, the ground-based cloud samples will be blocked, and then the traditional local binary pattern algorithm is optimized and the optimized algorithm of three different scales local binary pattern are used to extract the features of ground-based cloud images, and finally are used in the classification of the ground-based cloud. Experimental results show that multi-scale analysis method of gray scale and rotation invariant, reduces the effects of light and rotation distortion effects on the identification, through block processing to enhance the ability of local characterization, making it more suitable for ground-based cloud image recognition.