| With the deepening of the tourism information process,the construction of intelligent scenic spots has become an important goal for the development of scenic spots at home and abroad.Green energy traffic excursion boats are the main means of transportation for tourists to and from various attractions within the water scenic spots,and water traffic safety has become a key task in the management of scenic spots.At present,the scenic area mainly in the shorebased fixed position deployment monitoring means,cannot be timely control of the situation inside and outside the ship,so the realization of the scenic fleet video monitoring system has become an important research direction of water traffic safety control.In addition,for the image collected by shipboard camera in the water scenic areas,the rain and fog weather will cause clarity reduce,video defogging and image clarity processing becomes an important technical means to strengthen water traffic safety management in scenic areas,which is the focus of this thesis.Considering the problems above,this thesis takes the green energy traffic excursion fleet of a water wisdom scenic spot in North China as the application scenario,designs and implements the scenic fleet video monitoring system that can remotely play the monitoring screen inside and outside the ship,studies the image defogging method and applies it to the monitoring system.The research content of this thesis is as follows:(1)According to the construction requirements of the intelligent scenic spot,the video monitoring system of the scenic fleet based on the Internet of Things is designed.The system includes a shipboard subsystem on the ship side,a streaming media server and control command server on the cloud server side,and a shore-based subsystem on the monitoring side.(2)A deep learning-based image-defogging method is investigated.This thesis improves the defogging effect and efficiency of Light-Dehaze Net,a lightweight network model.To improve the defogging effect of the network model,the Re LU activation function in the model is replaced with Re LU6,the network layer is deepened and the attention mechanism module CBAM is introduced.To improve the defogging efficiency of the network model,the normal convolution in the model is replaced with a deep separable convolution.The experiments and related index evaluations of the network model before and after the improvement show that the above improvement methods have improved the defogging effect and efficiency of the model.(3)The implementation of the scenic fleet video monitoring system is completed.Firstly,based on the overall architecture of streaming media transmission,the RTP packets in the RTSP data stream output from the monitoring camera at the ship end are decapsulated to obtain video data.Then they encapsulated into FLV format and sent through RTMP protocol,completing the realization of the FFmpeg-based data pushing stream.Secondly,the construction and configuration of the Nginx-based streaming media server are completed,as the control command.Then,the VLC-based video playback software is designed at the monitoring end,and the streaming media playback,remote control of the streaming program,and the defogging function of the real-time stream are realized.Finally,the system-related tests are completed.Test results show that the system can achieve video playback latency of about 3 seconds,and real-time video streams to meet the needs of 25 fps defogging processing capacity.The scenic fleet video monitoring system has completed the system development,deployment,and testing in the first half of 2021.And it has been online in the scenic area running for a year,the system is stable and reliable,and has important significance to the scenic area water traffic safety management. |
PDF Full Text Request