Research On Rainfall Analysis Algorithm Of Marine Radar Based On Neural Network

Rainfall has always been one of the important factors in meteorological disasters,whether in the transportation industry or in daily life,sudden concentrated rainfall can cause major losses,accurate analysis of rainfall is of great significance in reducing rainfall disasters.With the development of meteorological technology,radar has become an important tool in rainfall detection,but weather radars are expensive and complex,and can only be constructed in key areas and cannot cover some remote and special areas.Now,marine radar has gradually become the first choice for rain measurement in local areas because of its low cost and sensitivity to rainfall.Therefore,this article chooses marine radar,use video echo data,combined with image processing and neural network technology to analyze local rainfall.The specific work content is as follows:(1)Build an X-band marine radar experiment platform.First,collect rainfall echo image,and then for the problem of ground clutter elimination,a method of thresholding segmentation and difference is applied,which effectively eliminates ground clutter and image noise,and a compensation algorithm is applied to eliminate The area is compensated;the electromagnetic wave attenuation design algorithm is compensated to obtain a more realistic echo image.Finally,an image processing display platform is designed to integrate processing algorithms and subsequent rainfall analysis functions to achieve the integration of image processing,area selection,rainfall analysis,image storage and other functions,provide experimental data for subsequent rainfall analysis,and make rainfall images The processing and analysis are more convenient and intuitive.(2)The research of artificial neural network in rainfall analysis algorithm.Aiming at the problem of rainfall analysis,an analysis algorithm from the perspective of echo intensity is proposed.Using artificial neural network,the rain echo intensity and radar receiver gain are considered as influencing factors,and the echo intensity,receiver gain-rainfall type data pairs are established.Then,BP neural network and RBF neural network are used to establish rainfall estimation model combined with data for training and recognition,and the model is applied to the built experimental display platform to analyze the rainfall in real time for the intensity of the video echo.The experimental results show that the method can effectively judge the rainfall type and has a good mapping relationship.At the same time,the RBF neural network is more suitable for rainfall analysis on the whole and can be applied to actual rainfall analysis.(3)Rainfall analysis algorithm based on convolutional neural network.Aiming at the problem of rainfall analysis,using the powerful capabilities of convolutional neural networks in the field of image recognition,an algorithm for rainfall analysis from the perspective of rainfall echo images is proposed.In order to obtain different levels of image features,combined with the idea of feature fusion,a simple model was designed and merged with the Alex Net model to improve it.At the same time,the inverted residual structure was introduced,and the Mobile Net v2 network model was used and improved.The rain image classification of heavy rain,moderate rain and light rain on the marine radar is used as the original image input to the network for training,and the recognition result is obtained.Experiments show that the feature fusion method can effectively help the network to obtain features,further improve the accuracy of rainfall analysis,meet the actual application requirements in rainfall analysis,and have advantages with rainfall analysis based on echo intensity,and can be effectively applied Various occasions.