基于图像质量分析的PM<sub>2.5</sub>空气质量预测

基于图像质量分析的PM_2.5空气质量预测

摘要: 为了提高空气污染物PM_2.5质量浓度预测的准确性，提出了一种基于图像数据预测PM_2.5质量浓度的方法.首先用手机或相机获取图像数据，然后用图像质量分析模型提取与PM_2.5质量浓度相关的特征向量作为输入，建立一个基于粒子群优化（particle swarm optimization，PSO）算法的支持向量回归机（support vector regression，SVR）（PSO-SVR）预测模型来估计PM_2.5的质量浓度.实验结果表明，与SVR模型和用遗传算法（genetic algorithm，GA）优化的支持向量回归机（GA-SVR）模型相比，PSO-SVR模型在预测准确性和实施效率方面具有更好的预测性能.

Abstract: To improve the prediction accuracy of air pollutants of the mass concentration of PM_2.5, a method of PM_2.5 mass concentration prediction based on collected image data were proposed. First, image data were acquired by mobile phones or cameras, and then feature vectors related to PM_2.5 mass concentration were extracted by image quality analysis model as input. A support vector regression (SVR) prediction model based on particle swarm optimization (PSO) algorithm (PSO-SVR) was established to estimate the mass concentration of PM_2.5. Results show that the prediction accuracy and efficiency of the PSO-SVR model are better than that of the SVR model and the support vector regression model optimized by genetic algorithm (GA-SVR).