The identification of hazardous segments is always the first step of a safety improvement program. However, the existing hazardous identification methods, such as the Crash Frequency method and the Empirical Bayes method, have a shortcoming in common: the process of road segmentation. The primary objective of this paper is to propose a new hazardous identification method based on the Kernel Density Estimation, in which the process of road segmentation is avoided and crash data is used in a new way. Crash data is collected on rural highways in Yangjiang in Guangdong and simulated data is generated from the crash data. The new method is compared with two commonly used hazardous segments identification methods using the simulated crash data. Quantitative evaluation tests proposed by prior researchers are used in the study of comparisons. There are two benefits of the new method: increasing the identification consistency and reducing false identification.
展开▼
