Identification method of traffic accident black spots based on dynamic segmentation and DBSCAN algorithm(PDF)
长安大学学报（自然科学版）[ISSN:1006-6977/CN:61-1281/TN]
- Issue:
- 2018年05期
- Page:
- 131-138
- Research Field:
- 交通工程
- Publishing date:
Info
- Title:
- Identification method of traffic accident black spots based on dynamic segmentation and DBSCAN algorithm
- Author(s):
- GENG Chao1; 2; PENG Yuhua1
-
- Keywords:
- traffic engineering; traffic safety; black spot; accumulative frequency curve method; equivalent accident number; DBSCAN algorithm
- PACS:
- -
- DOI:
- -
- Abstract:
- To improve identification accuracy of black spots in traffic accidents, an appropriate length of a road segment was selected as the basic unit for analysis and divided into several section combinations by the mobile step method. The mobile step method considered the number of traffic accidents and resulting casualties for a quantitative measure of traffic accidents. The accumulative frequency method was used to analyze the accidents, and obtain sections with relatively concentrated number of accidents under various combinations, then which were merged as the primary black spots according to the principle of adjacent sections. For primary black spots, appropriate neighborhood and threshold values were selected, and the densitybased spatial clustering of applications with noise (DBSCAN) algorithm was used for cluster analysis to seek out the “real” black spots with short lengths and concentrated accident occurrences. To verify the reliability of the identification method, black spots were identified for various sections of the Mazhao Expressway. The results show that length of the accidentprone section identified by the accumulative frequency method accounts for 28.4% of the total length of the segment, and the number of equivalent accidents accounts for 56.4% of the total number of accidents，on the basis of dynamic segment division. Thus, dividing the road unit dynamically by the mobile step method can identify the black spots to a great extent. The lengths of the real black spots only account for 10.8% of the total length, and the number of equivalent accidents accounts for 52.5% of the total number of accidents，When the DBSCAN algorithm is used to reexamine the obtained primary black spots. Thus the length of black spot sections have decreased by 61%, while the number of accidents remained almost the same. The combination of dynamic segmentation and the DBSCAN algorithm can be used to eliminate “nonblack” road sections in the primary road segment and identify the real black point segments. This method greatly improves the identification accuracy of black spots, and provides reliable technical support for effective elimination of traffic accident black spots. 3 tabs, 5 figs, 19 refs.
Last Update: 2018-10-23