Research On Hail Detection Method Based On Image Feature Analysis

Hail disasters bring huge economic losses to the people and the country every year,so the accurate identification of hail disasters plays an important role in the protection of people's property and national facilities.At present,researchers mostly use the analysis of the Doppler radar echo of hail clouds to realize the early warning and recognition of hail disasters,but there will still be certain errors in the recognition results.This paper proposes to detect the falling hail from the perspective of images,and use image processing technology as the posterior method of radar detection to further improve the accuracy of radar prediction and recognition.This article focuses on the hotspots and difficulties in image detection algorithms.The specific content is as follows:First,the collection equipment,collection method and collection environment of hail pictures are introduced.Then preprocess the collected hail pictures to eliminate possible noise,light and other factors.After comparing three color image filtering methods,choose the best distance direction filter to filter the image;when light and other factors When the image effect is not good,this article chooses the MSRCR algorithm to enhance the image quality.The preprocessing of the image can better increase the detailed information of the image and prepare for the subsequent feature extraction.Secondly,it designs and implements a hail recognition method that combines multiple features under the framework of D-S evidence theory.The color feature and texture feature of the preprocessed image are extracted,and a classification model described by support vector data is established for recognition.The experimental results show that the recognition effect of hail based on a single feature is not good.The D-S evidence theory is used to fuse color and texture features to identify hail.The results of comparative experiments show that the multi-feature fusion method based on D-S evidence theory can effectively improve the recognition rate of hail and reduce its false recognition rate and rejection rate.Finally,the characteristic parameters of the hail in the image are measured.According to the characteristics of hail samples,this paper proposes a segmentation method based on HSI color space and watershed algorithm.This method is based on the improvement of the traditional watershed algorithm.The improved watershed algorithm first performs the segmentation of the HSI color model,and then uses the distance change and the minimum value transformation to eliminate the redundant local minimum.For the separated hail particles,the statistical pixel method,the Freeman chain code and the improved minimum circumscribed rectangle method were used to measure the area,circumference and diameter of the hail particles.The experimental results show that the method used has high accuracy and good correlation.