基于AMOS观测和监控视频资料预测能见度

The Prediction of Visibility Based on AMOS Observation and Surveillance Video Data

雾的演化复杂多变,对交通运输和环境监测造成重大困扰,引起广泛关注.由于雾的演化规律受众多非线性因素的影响,难以准确预测其变化趋势.据此,构建了一种能够准确估计当前能见度、预测能见度变化趋势的模型.首先,建立了基于气象数据准确估计当前能见度的广义函数相加模型;其次,基于监控视频数据建立了具备高精度的VGG深度学习能见度估计模型;进而,建立了景深亮度差异的暗通道算法对能见度进行准确估计;最后,采用ARIMA模型对能见度进行预测,揭示了大雾的演化规律.

The evolution of fog is complex and changeable,which causes great troubles to transportation and environmental monitoring,and attracts widespread attention.Because the evolution law of fog is affected by many nonlinear factors,it is difficult to predict its changing trend accurately.Therefore,a model which can accurately estimate the current visibility and predict the changing trend of visibility is established in this paper.Firstly,a generalized additive modelis established to accurately estimate the current visibility based on meteorological data.Secondly,a VGG deep learning visibility estimation model with high accuracy is established based on surveillance video data.Furthermore,the dark channel algorithmis established to estimate the visibility accurately based on the difference of depth of field brightness.Finally,the ARIMA model is used to predict the visibility,and the evolution law of fog is revealed.