|
home / centre for automotive safety research / Publications / List / Details Publication Details| Title | Development of an Algorithm to Predict Pedestrian Injury Severity based on Data from the South Australian Traffic Accident Reporting System | | Authors | Nishimoto T, Mukaigawa K, Ponte G | | Year | 2015 | | Type | Conference Paper | | Abstract | Pedestrian crash data was extracted from the South Australian Traffic Accident Reporting System for the years 2000 to 2013. There were 5,000 crashes involving pedestrians during this period. A logistical regression model was constructed based on police reported speed and coded injury severity, to determine the effect of vehicle speed on probability of a serious or fatal pedestrian injury. In this paper we developed a base algorithm using vehicle speed to predict probability of injury. In addition to vehicle speed, a pedestrian age algorithm was also developed.
The prediction algorithm was developed in this study by using a logistic regression analysis. The algorithm predicts the probability of a pedestrian being ‘admitted to hospital/fatal’. The injury category ‘admitted to hospital/fatal’ was used as the severe injury threshold for the AACN system.
An example of the calculation of ‘admitted to hospital/fatal’ rate using the predictive algorithm for pedestrian injury is shown in figure 1 for a hypothetical vehicle speed of 40 km/h. In the base algorithm, the sum Z of the linear combination is calculated to be -0.679 and substituting this into equation gives the result P=33.7. Therefore, the ‘admitted to hospital/fatal’ rate of pedestrians under these conditions, according to the base algorithm is 33.7 percent.
By varying the vehicle speed and pedestrian age, the prediction algorithm for pedestrian injuries can be used to generate risk curves that show the relationship between speed, age and the probability of a pedestrian injury being ‘admitted to hospital/fatal’. Figure 2 shows the risk curve of for the base algorithm. Even at a vehicle speed of 0 km/h the ‘admitted to hospital/fatal’ rate exceeds 10%, and the rate increases proportionally with vehicle speed. The pedestrian algorithm based on Japanese crash data indicates that pedestrian death/serious injury rates for sedan type vehicles colliding with a 50 year old pedestrian at 40km/h is 29%. At the same vehicle speed in the present study based on SA crash data, the base prediction algorithm gives an ‘admitted to hospital/fatal’ rate of 34%, which is similar to the rate given by the algorithm based on Japanese data. | | Report Number | 20156114 | | Conference Name | Society of Automotive Engineers of Japan Annual Congress (Autumn) | | Conference Abbreviation | JSAE | | Conference Location | Fukuoka, Japan | | Conference Date | 14-16 October 2015 |
| Reference | | Nishimoto T, Mukaigawa K, Ponte G (2015). Development of an Algorithm to Predict Pedestrian Injury Severity based on Data from the South Australian Traffic Accident Reporting System. Society of Automotive Engineers of Japan Annual Congress (Autumn), Fukuoka, Japan, 14-16 October 2015. |
|
