基于熵值法的事故再现异常结果识别与融合

Identification and fusion of abnormal results in accident reconstruction based on entropy method

为客观融合多个事故再现结果并合理处理异常信息,提出一种基于熵值法的熵判断法以识别异常结果,并在融合时降低其概率。首先,熵判断法根据不同事故再现模型所得结果确定最终输出结果的子区间,计算各模型结果对各子区间的评价获得评价矩阵;其次,基于评价矩阵及信息熵获得各模型的信度,并预警模型信度最低者且认定为异常结果;然后,在确定无法剔除异常结果后,再结合评价矩阵及模型信度计算各子区间的概率,并在计算过程中降低异常结果的概率;最后,将所有子区间及其对应概率值作为最终融合结果输出。数值算例表明:当所有再现结果都可靠时,熵判断法所得融合结果与现有方法十分接近;而当结果中存在异常值时,熵判断法不仅能识别出异常结果且最终融合结果仍近似服从正态分布;真实汽车-摩托车事故证实该方法在实践中的实用性。

In order to objectively fuse multiple accident reconstruction results and handle abnormal information reasonably,an entropy-based entropy judgment method was proposed to identify abnormal results and reduce their probability during fusion.Firstly,the sub-intervals of the final output results were determined by the entropy judgment method based on the results obtained from different accident reconstruction models,and calculates the evaluation matrix of each model's results for each sub-interval.Secondly,based on the evaluation matrix and information entropy,the confidence of each model was obtained,and the model with the lowest confidence was warned and identified as the abnormal result.Then,after determining that abnormal results cannot be eliminated,the probability of each sub-interval was calculated based on the evaluation matrix and model confidence,and the probability of abnormal results was reduced during calculation.Finally,all sub-intervals and their corresponding probability values were output as the final fusion results.Numerical examples show that when all reconstruction results are reliable,the fusion results obtained by the entropy judgment method are very similar to existing methods.When there are abnormal values in the results,the entropy judgment method can not only identify abnormal results,but the final fusion results still approximately follow a normal distribution.A real car-motorcycle accident confirms the effectiveness of this method in practice.