Study And Implementation Of The Information Retrieval System For The Interaction Between Vessels And Signal Station In Controlled Waterways

Vessels can only traverse in single direction with respect to certain disciplinesprovided by signal stations due to the narrow and winding features of controlledwaterways as well as the limited viewing angles between vessels from upstream anddownstream. Hence, accurate decisions made by signal stations and the effect ofobedience to traffic rules contribute considerably to the traffic safety of controlled areas.In order to extract evidences of accidents and retrieve “points of interest” from massivemulti-domain sensing information, research is motivated to rebuild the interactionbetween behaviors of vessels and decisions of signal stations based on the history data.Based on the realistic data collected from the Lu-yu section of Yangtze River, thisthesis presents a novel algorithm to organize and display multi-domain sensinginformation by leveraging their time-dependent characteristics. An information retrievalsystem is developed to reconstruct the scenes in the controlled waterway in any specifictime interval. The main work of the thesis and achievements are summed up as follows:①Analyze the characteristics and storage modes of the multi-source data(decision data, AIS data, surveillance videos and VHF audio data) gathered during theinteraction of signal stations and vessels to build exhaustive data description frameworkwith unified format.②Conduct index-related search by making use of the overlapping and sequentialfeatures of multi-source data and display tracks of vessels on the electronic map in areal time manner. Moreover, the interaction reality between vessels and signal stations isrebuilt by using associated audios and videos.③Generate a time series model by extracting characteristics in the controlledwaterway (e.g. the number of vessels, the direction of ships, the traversing time in thecontrolled waterway) and present a dynamic time wrapping (DTW) based algorithm torealize interesting data locating and mining. On top of this algorithm, themultimedia-on-demand service is realized to satisfy the specific requirement of diverseusers.By employing the C++programming language, this thesis develops an informationretrieval system that can rebuild history scenes in a controlled waterway based on themulti-domain sensing information. The system mainly concentrates on reconstructingthe interaction between commands of signal stations and movements of vessels, and can effectively improve the efficiency of data management in a controlled waterway andoptimize the decision-making process.