Design And Implementation Of Information Supervision System For Ferry In Luzhou Section Of Yangtze River

Most of the ferries in Luzhou section of the Yangtze River operate in rural towns and villages,which are characterized by complex channel,weak ferry infrastructure,small ferry size,poor boat condition and poor ferry workers’ comprehensive quality.At the same time,the annual ferry transport passengers are about 1 million,Once a dangerous situation occurs in the course of ferry sailing,it is very easy to cause a major life safety accident,resulting in serious consequences.At present,the supervision of ferries in Luzhou section of the Yangtze River basically adopts the mode of artificial on-site supervision,and adopts the traditional tactics of patrol,marking and garrison.The application rate of information technology,risk management and control level and supervision efficiency are low.For the Luzhou section of the Yangtze River ferry supervision work in the emergence of information supervision needs,this paper designs and realizes a kind of video monitoring technology,the convolution neural network technology,the deep learning target detection technology for maritime regulators ferry information supervision system,realize real-time video surveillance for the ferry passenger and the ferry passengers life jacket wearing behavior detection.The system is proved to be able to meet the requirements of maritime regulatory authorities after the actual ship test.The main research contents of this paper are as follows:(1)Study and analyze the supervision status quo of ferry in Luzhou section of the Yangtze River,find out the problems existing in the actual supervision work,clarify the problems that need to be solved in the information supervision system,put forward the design requirements of the system,design the overall architecture of the system on this basis,and study the operation process of key functional modules of the system.(2)Based on the convolutional neural network technology,the passenger life vest detection algorithm is studied.Because the detection of life jacket needs to identify the life jacket and mark its position on the image,the algorithm needs to choose from the target detection algorithm.After further studying the RCNN series algorithm and YOLO series algorithm,which are widely used and mature in technology at the present stage,according to the actual needs and experimental conditions of this paper,YOLO series algorithm is selected as the algorithm model of this paper,and the algorithm is studied in depth.(3)Images of passengers wearing life jackets were obtained through the Internet and surveillance cameras installed on the ferry,and Label Img software was used to label the collected images and produce training data sets that met the requirements.In view of the small amount of data in the data set of this paper,the transfer learning method was used to train the YOLOV3 life jacket detection model.After the training,the model was tested in real scene,and the indicators of the test results were analyzed.After solving the existing problems,the test was carried out again,and the final indicators of the test results met the requirements.(4)PyQT5 was used to encapsulate the YOLOV3 life jacket detection model and generate the desktop client system.The video monitoring system and life jacket detection system were integrated on the computer to build the Luzhou section of the Yangtze River ferry information supervision system,and the system functions were tested,and the test results met the requirements of Luzhou Maritime Safety Administration.