The Research On Information Modelling And Active Detection Method For Flood Events

Flood events frequently occurred worldwide,resulting in large numbers of casualties,serious economic losses,and huge environmental damages.The study of the occurrence and development law of flood events is not only great demands from the protection of people's lives and properties,the promotion of sustainable development of society and economy,the maintenance of national prosperity and stability and the construction of a harmonious society,but also the key for the promotion of applying the techniques of remote sensing and earth observation in the area of flood applications.But floods do not occur instantaneously,and the development of them often go through multiple phases of incubation,preparedness,occurrence,and recover,etc.During the full lifecycle of floods,from occurrence to the end,cross-sectoral dynamic and static information sharing,accurate extraction of the key factors of floods,such as the flood area,and the early detection and alert publishing of flood events are of great significance to increasing the flood responding efficiency,providing information support for flood decision-making,and mitigating or avoiding losses induced by floods.However,the existing studies could not yet solve these problems.Therefore,this paper carries out researches from the aforementioned three aspects,with the expectation of efficient flood event information sharing across departments,the accurate estimation of flooding areas,and the detection and notification of flood events in advance.The inter-departmental dynamic and static information sharing can promote the information exchange among the departments involved in the flood responding,reduce the cost of flooding information exchange,and improve the efficiency of flood coping.However,a unified dynamic and static flood event information representation model was absent,constituting obstacles to the flood information exchange.In order to solve the aforementioned problems,this paper analyzed the static and dynamic information sharing requirements of flood events,constructed the metamodelling framework,designed the basic metadata components and the concrete description structure,and proposed the object-oriented full lifecycle flood event information dynamic representation metamodel(FLFM),of which the extension mechanism and formalisation were meanwhile provided.Compared with the existing flood event information representation models,FLFM proposed in this paper could realize the unified dynamic and static information representation for different types of floods,satisfying the needs of cross-domain flood event information sharing and management.Flooding area is an important input during flood related decision-making.The accurate extraction of the flooding area could provide important decision information support for the flood coping,but the boundary of flood water body is often complicated with many islands inside,the extraction of which could not be implemented by the exsiting water boundary extraction methods.To tackle this problem,this paper analyzed the shortcomings of the existing balloon snake model,added the three mechanisms of adaptive image preprocessing,topology collision detection and processing,and the automatic termination judgment of contour iteration,empowering the balloon snake model the capability of weak boundary extraction,curve splitting and merging,and automatic termination.The experiments and the corresponding accuracy evaluation results demonstrated that the method proposed in this paper can be applied in the accurate boundary extraction of water bodies with inner islands,and thus can be utilized for the accurate estimation of flooding areas.Compared with the passive response,the early detection and alert releasing of flood events could spare valuable time for life saving and loss reduction,but the current detection method only rely on single threshold,the reliability and extensibility of which cannot be assured.This paper analyzed the relationship between historical observations and flood events,determined the factor most relevant to flood events and its data variation pattern during the flood event occurrence time,based on which the flood event filtering model was constructed,and proposed the active detection and alert releasing service method for flood events to solve this problem.The real-time observation data was accessed,the abnormal data was actively filtered,and the phase-based detection and alert releasing of flood events were achieved.Experiments demonstrated that the service method can discover the abnormal observation of the ground observations and timely release the warning message.Besides,a flood event information modeling and active detection prototype was designed and implemented in this paper,based on the flood event information representation model,the flooding area accurate extraction method,and the real-time active detection and alert release service of flood events.The historical flood event in Liangzi Lake was used as an example,and its information modeling,accurate flooding area estimation,and active detection and alert releasing were all implemented.The innovation of this paper is mainly reflected in three aspects:1)The observation-oriented full life cycle flood event information dynamic representation metamodel was proposed,and the unified modeling and full life cycle information management of flood events were realized.The effective description for the dynamic observation data of flood events laid model foundation for intergrated management and sharing service for the flood event information;2)An accurate boundary extraction method for water bodies with inner islands was proposed,which realized the accurate boundary extraction for water bodies with inner islands without the support of grid decomposition.The method could meanwhile ensure the accuracy and efficiency,providing method support for accurate boundary extraction of the flooding area with complicated borders and multiple innder islands;3)An active flood event detection and alert release service method was proposed,with the in-situ sensor observation accssed and filtered in real-time,the active flood event detection and alert releasing implemented,reserving preparation time for flood events,and reducing the possible losses caused by flood events.However,there are still many shortcomings in the construction of the task library and knowledge base of different types of flood events,the construction of flood event filtering model in large-scale watershed,and the simultaneous boundary extraction of multiple water bodies.These problems will be paid more attention to and can be expected to be gradually resolved in the future studies.