Research On Air Quality Measurement Method Based On Visual Information

Environmental air quality affects people’s lives all the time,it has certain reference value for people’s travel,work and other activities.In the task of air quality protection,obtaining real-time accurate and reliable ambient air quality is one of the important links.At present,the measurement of environmental air quality mainly adopts the method of setting up air quality monitoring stations in specific geographical locations of the city,and using air quality detectors to periodically sample and analyze the surrounding atmospheric environment.This method is easily limited by specific times and locations.In order to solve this problem,it is a practical idea to use mobile camera equipment to collect environmental images of the current area in real time,and use deep learning image recognition technology for air quality measurement.At present,some existing air quality measurement algorithms related to deep learning use a single convolutional neural network to extract the features of the entire image.However,due to the different features of each component of the environmental image,this method will ignore the differences of each part of the image.In order to solve the problem of feature extraction of different components of the environmental image,this paper proposes a double-channel weighted convolution network ensemble learning method.This method can achieve the feature extraction of different components of the environmental image and perform fine-quality air quality measurement.In addition,a feature weight self-learning method is proposed to find the optimal weight ratio of two parts of features.Experiments show that the algorithm can complete the tasks of air quality level measurement and air quality index measurement based on environmental images,and has achieved considerable accuracy in level measurement,which can be further applied to engineering practice.