In recent years,the degree of global environmental pollution has been increasing,and most cities in my country have also experienced varying degrees of air pollution.At present,the method of obtaining air quality data in my country is mainly to set up monitoring stations in cities to determine the concentration of pollutants,and then convert them into air quality index for residents to query.However,the cost of the equipment required by this method is relatively high,and its detection mode is coarse-grained and cannot be refined to every corner of the city.Now it is an informationized world.Compared with text information,image information is more intuitive and efficient.Currently,the image acquisition technology of mobile devices is constantly improving.Using images to detect air quality grades can make air quality detection fine-grained,low-cost,and has high research value.The thesis takes image-based air quality grade evaluation as the research object.First,explore and analyze the effects of air pollution on images through experiments.Analyze the characteristics of particulate matter in air pollution and summarize the differences in characteristics of images with different degrees of air pollution from the aspects of contrast difference characteristics,color space distribution characteristics and transmittance texture characteristics.Secondly,an air quality grade evaluation method based on image pixel ratio is proposed.This method mainly starts with the influence of air pollution on the saturation and Intensity channels in the image HSI space,and calculates the edge pixel ratio,high brightness pixel ratio,low saturation pixel ratio and high energy pixel ratio of the image in the HSI space as features.Combine the support vector machine to build an air quality grade evaluation model.The performance and applicability of this method are proved through different data sets and comparison with other methods.Experimental results show that this method can evaluate the air quality grade during image collection more accurately than other methods,and has the advantages of high accuracy,fast processing speed,low cost,good real-time performance,and easy promotion.Finally,designed and implemented an image-based air quality rating system.This system uses Java Web related technology combined with Open CV and Lib SVM tools to complete the four functions of image upload,image feature extraction,air quality grade evaluation model construction and air quality grade evaluation.Finally,test the system in terms of compatibility and load.The results show that the system can accurately assess the air quality grade at the time of image collection,which is convenient for people to better monitor the air quality of their environment and arrange their daily schedules accordingly. |