Jitpinun Piriyataravet, W. Kumwilaisak, J. Chinrungrueng
{"title":"基于深度神经网络和双向LSTM的公交车站自动检测","authors":"Jitpinun Piriyataravet, W. Kumwilaisak, J. Chinrungrueng","doi":"10.1109/ICA-SYMP50206.2021.9358448","DOIUrl":null,"url":null,"abstract":"This paper presents a novel method in bus stop prediction from bus GPS trajectories. Our proposed bus stop prediction algorithm is based on the deep neural network and time filtering algorithm. Bus speed histograms of all locations along a route are first constructed. A bus speed histogram and a bus heading direction at each location are input features of a deep neural network. A deep neural network consists of the CNN networks and fully connected networks. The outputs from a deep neural network of all locations along a route are inputs to the LSTM network. It outputs soft decisions of bus stop prediction of all locations. The time filtering algorithm refines the results obtained from the LSTM network. It constructs time histograms of all locations and extracts the most probable timestamps of all locations. Then, a linear regression method is used to correct timestamps. Time distributions can be derived from the updated timestamp and are compared with a reference distribution. Locations with time distributions close to the reference distributions are predicted as bus stop locations. We compare our algorithm on a set of GPS data of NSTDA bus service. The proposed technique can outperform conventional bus prediction methods.","PeriodicalId":147047,"journal":{"name":"2021 Second International Symposium on Instrumentation, Control, Artificial Intelligence, and Robotics (ICA-SYMP)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Automatic Bus Stop Detection with Deep Neural Networks and Bi-directional LSTM\",\"authors\":\"Jitpinun Piriyataravet, W. Kumwilaisak, J. Chinrungrueng\",\"doi\":\"10.1109/ICA-SYMP50206.2021.9358448\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"This paper presents a novel method in bus stop prediction from bus GPS trajectories. Our proposed bus stop prediction algorithm is based on the deep neural network and time filtering algorithm. Bus speed histograms of all locations along a route are first constructed. A bus speed histogram and a bus heading direction at each location are input features of a deep neural network. A deep neural network consists of the CNN networks and fully connected networks. The outputs from a deep neural network of all locations along a route are inputs to the LSTM network. It outputs soft decisions of bus stop prediction of all locations. The time filtering algorithm refines the results obtained from the LSTM network. It constructs time histograms of all locations and extracts the most probable timestamps of all locations. Then, a linear regression method is used to correct timestamps. Time distributions can be derived from the updated timestamp and are compared with a reference distribution. Locations with time distributions close to the reference distributions are predicted as bus stop locations. We compare our algorithm on a set of GPS data of NSTDA bus service. The proposed technique can outperform conventional bus prediction methods.\",\"PeriodicalId\":147047,\"journal\":{\"name\":\"2021 Second International Symposium on Instrumentation, Control, Artificial Intelligence, and Robotics (ICA-SYMP)\",\"volume\":\"37 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2021-01-20\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2021 Second International Symposium on Instrumentation, Control, Artificial Intelligence, and Robotics (ICA-SYMP)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICA-SYMP50206.2021.9358448\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 Second International Symposium on Instrumentation, Control, Artificial Intelligence, and Robotics (ICA-SYMP)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICA-SYMP50206.2021.9358448","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Automatic Bus Stop Detection with Deep Neural Networks and Bi-directional LSTM
This paper presents a novel method in bus stop prediction from bus GPS trajectories. Our proposed bus stop prediction algorithm is based on the deep neural network and time filtering algorithm. Bus speed histograms of all locations along a route are first constructed. A bus speed histogram and a bus heading direction at each location are input features of a deep neural network. A deep neural network consists of the CNN networks and fully connected networks. The outputs from a deep neural network of all locations along a route are inputs to the LSTM network. It outputs soft decisions of bus stop prediction of all locations. The time filtering algorithm refines the results obtained from the LSTM network. It constructs time histograms of all locations and extracts the most probable timestamps of all locations. Then, a linear regression method is used to correct timestamps. Time distributions can be derived from the updated timestamp and are compared with a reference distribution. Locations with time distributions close to the reference distributions are predicted as bus stop locations. We compare our algorithm on a set of GPS data of NSTDA bus service. The proposed technique can outperform conventional bus prediction methods.