Deep neural network model construction method, system and device for video fog monitoring

Abstract

The invention provides a deep neural network model construction method, system and device, and the method comprises the following steps: constructing a video fog recognition data set which comprises fog video data and fog-free video data; extracting single-frame images of the video data to form a sample set, and dividing the sample set into a training set and a test set; constructing a deep neural network model; training a deep neural network model by using the training set, and optimizing model parameters; and testing the trained deep neural network model by using the test set. According to the invention, the static features of the single-frame image can be extracted, the structure is simple, the recognition accuracy is high, real-time smoke fog monitoring recognition under the monitoring video can be realized, and the requirements of non-meteorological users on fog recognition business are met. In daytime and night heavy fog identification under various visibility levels, the average daytime identification rate is 85.4% or above, the average night identification rate is 69.7%, and the result can basically meet the requirement of the camera for assisting in monitoring the heavy fog level.

Description

technical field [0001] The invention belongs to the technical field of meteorological monitoring, and in particular relates to a method, system and device for constructing a deep neural network model for video heavy fog monitoring. Background technique [0002] Fog weather is an important disastrous weather that affects expressway safety, and improving the monitoring technology level of fog weather is an important measure to ensure traffic safety. In recent years, monitoring heavy fog by deploying automatic visibility stations along highways has played an important role in ensuring traffic safety. However, the automatic visibility stations are generally far away, more than 10km, and cannot provide monitoring and early warning services for small-scale local heavy fog and group fog. These small-scale group fog often easily lead to driver negligence or evasion, causing serious traffic accidents. ACCIDENT. [0003] With the development of monitoring level and informatization, ...

AI Technical Summary

Problems solved by technology

When the traffic control department judges the visibility level of heavy fog, it mainly checks the road weather conditions through manual inspection. This method is time-consuming, laborious and low-frequency, and it is difficult to detect severe weather conditions in time.

Images