{"title":"基于主动学习的道路网络分段匹配迭代框架","authors":"Wenhao Yu, Mengqi Liu","doi":"10.1080/15230406.2023.2190935","DOIUrl":null,"url":null,"abstract":"ABSTRACT Road network matching that detects arc-to-arc relations is a crucial prerequisite for the update of road data. The increasing complexity of multi-source and multi-scale road network data challenges the existing methods on accuracy and efficiency. This paper focuses on the interactive-based probabilistic relaxation approach. It is difficult to obtain satisfactory results by using completely automatic matching algorithm in some complicated road networks such as multi-lane carriageways. We try to improve the matching accuracy by combining optimization matching model with manual interaction. The method uses the module of active learning to construct unlabeled sample pool from preliminary matching of probabilistic relaxation, and then selects the arcs with the highest uncertainty by query function. The selected road is then handed over to humans to determine its arc-to-arc relations in the other road network. Finally, the matching parameters are automatically adjusted according to the user’s feedback information, so as to realize the dynamic optimization of the model. Our interaction method is efficient as it only needs to specify few arc-to-arc mappings and others can be amended automatically. Our experimental results reveal that active learning can substantially improve the performance of standard probabilistic relaxation algorithms in road network matching.","PeriodicalId":47562,"journal":{"name":"Cartography and Geographic Information Science","volume":"50 1","pages":"333 - 350"},"PeriodicalIF":2.6000,"publicationDate":"2023-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"An iterative framework with active learning to match segments in road networks\",\"authors\":\"Wenhao Yu, Mengqi Liu\",\"doi\":\"10.1080/15230406.2023.2190935\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"ABSTRACT Road network matching that detects arc-to-arc relations is a crucial prerequisite for the update of road data. The increasing complexity of multi-source and multi-scale road network data challenges the existing methods on accuracy and efficiency. This paper focuses on the interactive-based probabilistic relaxation approach. It is difficult to obtain satisfactory results by using completely automatic matching algorithm in some complicated road networks such as multi-lane carriageways. We try to improve the matching accuracy by combining optimization matching model with manual interaction. The method uses the module of active learning to construct unlabeled sample pool from preliminary matching of probabilistic relaxation, and then selects the arcs with the highest uncertainty by query function. The selected road is then handed over to humans to determine its arc-to-arc relations in the other road network. Finally, the matching parameters are automatically adjusted according to the user’s feedback information, so as to realize the dynamic optimization of the model. Our interaction method is efficient as it only needs to specify few arc-to-arc mappings and others can be amended automatically. Our experimental results reveal that active learning can substantially improve the performance of standard probabilistic relaxation algorithms in road network matching.\",\"PeriodicalId\":47562,\"journal\":{\"name\":\"Cartography and Geographic Information Science\",\"volume\":\"50 1\",\"pages\":\"333 - 350\"},\"PeriodicalIF\":2.6000,\"publicationDate\":\"2023-04-11\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Cartography and Geographic Information Science\",\"FirstCategoryId\":\"89\",\"ListUrlMain\":\"https://doi.org/10.1080/15230406.2023.2190935\",\"RegionNum\":3,\"RegionCategory\":\"地球科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"GEOGRAPHY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Cartography and Geographic Information Science","FirstCategoryId":"89","ListUrlMain":"https://doi.org/10.1080/15230406.2023.2190935","RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"GEOGRAPHY","Score":null,"Total":0}
An iterative framework with active learning to match segments in road networks
ABSTRACT Road network matching that detects arc-to-arc relations is a crucial prerequisite for the update of road data. The increasing complexity of multi-source and multi-scale road network data challenges the existing methods on accuracy and efficiency. This paper focuses on the interactive-based probabilistic relaxation approach. It is difficult to obtain satisfactory results by using completely automatic matching algorithm in some complicated road networks such as multi-lane carriageways. We try to improve the matching accuracy by combining optimization matching model with manual interaction. The method uses the module of active learning to construct unlabeled sample pool from preliminary matching of probabilistic relaxation, and then selects the arcs with the highest uncertainty by query function. The selected road is then handed over to humans to determine its arc-to-arc relations in the other road network. Finally, the matching parameters are automatically adjusted according to the user’s feedback information, so as to realize the dynamic optimization of the model. Our interaction method is efficient as it only needs to specify few arc-to-arc mappings and others can be amended automatically. Our experimental results reveal that active learning can substantially improve the performance of standard probabilistic relaxation algorithms in road network matching.
期刊介绍:
Cartography and Geographic Information Science (CaGIS) is the official publication of the Cartography and Geographic Information Society (CaGIS), a member organization of the American Congress on Surveying and Mapping (ACSM). The Cartography and Geographic Information Society supports research, education, and practices that improve the understanding, creation, analysis, and use of maps and geographic information. The society serves as a forum for the exchange of original concepts, techniques, approaches, and experiences by those who design, implement, and use geospatial technologies through the publication of authoritative articles and international papers.