基于多源证据的繁忙水域交管雷达异常目标识别方法研究

项目名称： 基于多源证据的繁忙水域交管雷达异常目标识别方法研究

项目编号： No.61503289

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 马枫

作者单位： 武汉理工大学

项目金额： 20万元

中文摘要： 繁忙水域下，交管雷达由于多路径效应、航道建筑物的影响，返回大量真伪不分、重要性不明的跟踪目标，依赖人工识别，无法实现自动管控。随着共享障碍的消除，使得利用船舶自动识别系统（AIS）、签证、事故记录等多源信息挖掘统计特征，识别出错误、高风险等异常目标成为可能。研究首先根据真伪、风险等特征，建立幂集上的识别框架，分析雷达传感与多源信息的不确定性，建立特征相关的多源证据；然后，基于证据推理规则（Evidential Reasoning Rule）合成证据，建立识别模型；最后以人工经验拟合、历史数据趋近等多重优化目标，在验证样本或人工干预的支持下，设计训练过程，实现模型参数、结构自整定。重点解决异构数据特征映射、不确定性度量与取值、先验概率稳定性评估等难点问题。旨在利用多源历史数据建立的先验统计信息，研究符合船舶特征的历史分布、人工经验的雷达目标特征智能识别方法，为提高监管效率、改善海事安全服务。

中文关键词： 水运交通；交通状态感知；不确定性分析；多目标优化；证据推理

英文摘要： In busy waterway, due to the multipath effect and block by channel buildings, there are large number of targets with unspecified authenticity and importance obtained by maritime radars, which relies on the manual judgment, increasing the burden of supervisors. By sharing the Automatic Identification System, endorsement record, accident notes, multi-source information, it is possible to mining the characteristic in specific record, and recognize the fault and high risk targets. In the first step, the research proposes the discrimination framework on the power set, analyzes the uncertainty of sensor and source information, builds up the evidences on multi-source characteristic. Secondly, construct the recognition process on Evidential Reasoning Rule, ER Rule. At last, with the human experience satisfaction and historical data proximity, build the multi-goal optimization model, which is able to train the parameters and structure by learning the small sample or limited manual intervene. The core problems are the method to map the multi-source heterogeneous to characteristic prior probability, how to measure and evaluate the uncertainty, and estimate the stability of prior probability. With this method research, construct a stable module which is able to recognize the feature of specific radar target, by analyzing the multisource historical data and human experience, to increase the efficiency, improve the safety.

英文关键词： waterway transportation;traffic perception;uncertainty analysis;multi-object optimization;Evidential Reasoing