Design And Realization Of Short-term Cloud Images Forecasting System Based On OpenCV

Weather forecasts are closely related to people's daily lives,for example,people's daily travel,activities,disaster prevention and mitigation,not only to ensure the smooth development of the economy but also for the people's peace and happiness made a great contribution.Withing the development of science and technology,meteorological monitoring tools also developed rapidly,because of the rich information,weather images are widely used.By combining meteorological cloud images and high-performance computers,making predictions of cloud images more realistic.First of all,this paper introduces the current research on cloud detection and tracking at home and abroad,and then analyzes and studies the common methods of tracking and forecasting moving objects.Although the study of short-term cloud prediction in recent years has never been interrupted,it also have caused some difficulties due to the non-rigid body and the characteristics of mutation.Therefore,this paper combines the actual situation and system requirements,the proposed pyramid optical flow combined with discrete kalman filter ideas to predict the cloud image.The weather images forecasting system in this paper mainly focuses on three steps:The first step is the preprocessing process of the meteorological cloud,after analyzing the characteristics of the cloud image,the median filter algorithm is used to remove the noise from the cloud image and can provide accurate data for the subsequent work.The second step is to extract the motion vector of the cloud,the horn-schunck constraint equation in the optical flow method can make the algorithm have global smoothness,and then combine the pyramid layering technology to reduce the iterative calculation in the optical flow operation,greatly reducing the calculation of time.The third step is to track the cloud and to complete the prediction of the cloud,introduces a discrete kalman filter,it is able to achieve more precise positioning cloud and have a better implementation of cloud forecasting.The system able to obtain data from the local and combining the abundant information of the meteorological cloud images can be more accurate to predict thenext moment of the cloud.