Automatic Traffic Surveillance Using Deep Learning Model Under Extreme Climatic Conditions

This research makes use of deep learning for creating an automatic traffic surveillance model to detect the various vehicles under extreme climatic weather conditions. The model proposed is an extension of the standard VGG16-Net model. The model has been tested using real time videos taken from online database repository. The dataset is available in the form of images of vehicles running under three conditions namely normal, snowy and rainy conditions. The model detects the conditions of weather in the traffic when vehicle move on road. Weather conditions namely rainy and snowy are predicted in this research. The model has been developed using Python programing language. Algorithms adopted in this research are Adam Optimization Algorithm (AOA) and Stochastic Gradient Descent (SGD) for optimizing the model. Performance is estimated through f1-score, precision, accuracy and recall parameters. It was found that model for predicting weather condition showed 92 percent accuracy