The Research On Traffic Incidents Detection Based On Hybrid Multi Support Vector Machine

Along with society fast progressive and highway's development, on highway's transportation event occurs frequently, serious influence social economy development. Therefore, various countries have carried out the highway event automatic detection algorithm research in abundance, carries on the highway transportation event's automatic detection. (MSVM-AID) the algorithm appears based on the support vector machines multi-sorter's transportation event detection, under the success solution small sampled data condition, the model training the study with has owed the study phenomenon, and can make the good response to the transportation event's occurrence. But, we discovered in the practice research, the MSVM-AID algorithm in examines in certain particular cases the transportation event, if regarding traffic lane number many highways, once has the transportation event, and the event vehicles take the traffic lane to be limited, hands over in the throughflow parameter fluctuation not big situation, its examination precision is not very ideal. Therefore, this article in view of this question, makes the improvement to the MSVM-AID algorithm model, simultaneously has introduced the series, the parallel two sorter combined method, designed mixed the MSVM-AID algorithm. Meanwhile carries on the redesign to the eigen vector/feature vector/proper vector, constructed three to have the method effect eigen vector/feature vector/proper vector, enlarged hands over the throughflow parameter in the multiple track small transportation event's change degree. This algorithm divides into parallel and the series connected two parts, to the small transportation event, makes 2 examination analyses, solved the MSVM-AID algorithm to leak the sentencing rate high question, simultaneously has promoted the model whole examination effect and the portability.