GNSS photo matching: Positioning using gnss and camera in urban canyon

Taro Suzuki, Nobuaki Kubo

研究成果: Conference contribution

1 引用 (Scopus)

抄録

This paper describes a positioning technique that can be used for identifying pedestrians in urban environments in which there are large global navigation satellite system (GNSS) positioning errors because of poor satellite geometry and multipath signals. We propose a technique to realize position estimation based on GNSS and photos taken with a smartphone. We combine GNSS with a 3D map, camera, accelerometer, and magnetic sensor on a smartphone to improve positioning accuracy in urban environments. The main objective of the proposed method is to reduce the GNSS position error distribution, which is biased in the direction pointing across the street because of signal blocking in urban canyons. First, we calculate a positioning solution and its distribution using normal GNSS point positioning with a strict signal strength threshold and high-elevation mask angles. In GNSS positioning results, the horizontal position error distribution typically extends across the street. Multiple position candidates near the true position are then generated from the GNSS position error distribution. Next, we capture a photo using a smartphone that is aimed at the sky. The likelihood of the position candidates are then determined based on an image-matching technique that uses the captured photo and a virtual photo, which is simulated based on the candidate position, camera attitude, and 3D map. We use Google Earth to generate the virtual photo. Finally, the position candidate that has the highest likelihood is chosen as the final estimated user position. Using this technique, the error distribution of GNSS positioning is corrected by the camera, accelerometer, magnetic sensor, and 3D map. Thus, a more accurate user position can be determined. To confirm the effectiveness of the proposed technique, a positioning experiment was performed in a real-world urban-canyon environment. We used a GNSS receiver and a smartphone to collect data from multiple locations and evaluated the accuracy of the proposed technique. In this experiment, similar virtual photos can be generated at the true position from a 3D map. In other words, the likelihood of a position candidate increases when it is near the true position. In conclusion, our proposed method is effective for estimating accurate user positions in urban canyons, where conventional GNSS has large positioning errors.

元の言語English
ホスト出版物のタイトル28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015
出版者Institute of Navigation
ページ2470-2480
ページ数11
4
ISBN(電子版)9781510817258
出版物ステータスPublished - 2015
イベント28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015 - Tampa, United States
継続期間: 2015 9 142015 9 18

Other

Other28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015
United States
Tampa
期間15/9/1415/9/18

Fingerprint

Navigation
Cameras
Satellites
Smartphones
Magnetic sensors
Accelerometers
Image matching
Normal distribution
Masks
Earth (planet)
Experiments
Geometry

ASJC Scopus subject areas

  • Computer Science Applications
  • Electrical and Electronic Engineering
  • Software

これを引用

Suzuki, T., & Kubo, N. (2015). GNSS photo matching: Positioning using gnss and camera in urban canyon. : 28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015 (巻 4, pp. 2470-2480). Institute of Navigation.

GNSS photo matching : Positioning using gnss and camera in urban canyon. / Suzuki, Taro; Kubo, Nobuaki.

28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015. 巻 4 Institute of Navigation, 2015. p. 2470-2480.

研究成果: Conference contribution

Suzuki, T & Kubo, N 2015, GNSS photo matching: Positioning using gnss and camera in urban canyon. : 28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015. 巻. 4, Institute of Navigation, pp. 2470-2480, 28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015, Tampa, United States, 15/9/14.
Suzuki T, Kubo N. GNSS photo matching: Positioning using gnss and camera in urban canyon. : 28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015. 巻 4. Institute of Navigation. 2015. p. 2470-2480
Suzuki, Taro ; Kubo, Nobuaki. / GNSS photo matching : Positioning using gnss and camera in urban canyon. 28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015. 巻 4 Institute of Navigation, 2015. pp. 2470-2480
@inproceedings{f63f5dd77fc34788988ad034377537e6,
title = "GNSS photo matching: Positioning using gnss and camera in urban canyon",
abstract = "This paper describes a positioning technique that can be used for identifying pedestrians in urban environments in which there are large global navigation satellite system (GNSS) positioning errors because of poor satellite geometry and multipath signals. We propose a technique to realize position estimation based on GNSS and photos taken with a smartphone. We combine GNSS with a 3D map, camera, accelerometer, and magnetic sensor on a smartphone to improve positioning accuracy in urban environments. The main objective of the proposed method is to reduce the GNSS position error distribution, which is biased in the direction pointing across the street because of signal blocking in urban canyons. First, we calculate a positioning solution and its distribution using normal GNSS point positioning with a strict signal strength threshold and high-elevation mask angles. In GNSS positioning results, the horizontal position error distribution typically extends across the street. Multiple position candidates near the true position are then generated from the GNSS position error distribution. Next, we capture a photo using a smartphone that is aimed at the sky. The likelihood of the position candidates are then determined based on an image-matching technique that uses the captured photo and a virtual photo, which is simulated based on the candidate position, camera attitude, and 3D map. We use Google Earth to generate the virtual photo. Finally, the position candidate that has the highest likelihood is chosen as the final estimated user position. Using this technique, the error distribution of GNSS positioning is corrected by the camera, accelerometer, magnetic sensor, and 3D map. Thus, a more accurate user position can be determined. To confirm the effectiveness of the proposed technique, a positioning experiment was performed in a real-world urban-canyon environment. We used a GNSS receiver and a smartphone to collect data from multiple locations and evaluated the accuracy of the proposed technique. In this experiment, similar virtual photos can be generated at the true position from a 3D map. In other words, the likelihood of a position candidate increases when it is near the true position. In conclusion, our proposed method is effective for estimating accurate user positions in urban canyons, where conventional GNSS has large positioning errors.",
author = "Taro Suzuki and Nobuaki Kubo",
year = "2015",
language = "English",
volume = "4",
pages = "2470--2480",
booktitle = "28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015",
publisher = "Institute of Navigation",
address = "United States",

}

TY - GEN

T1 - GNSS photo matching

T2 - Positioning using gnss and camera in urban canyon

AU - Suzuki, Taro

AU - Kubo, Nobuaki

PY - 2015

Y1 - 2015

N2 - This paper describes a positioning technique that can be used for identifying pedestrians in urban environments in which there are large global navigation satellite system (GNSS) positioning errors because of poor satellite geometry and multipath signals. We propose a technique to realize position estimation based on GNSS and photos taken with a smartphone. We combine GNSS with a 3D map, camera, accelerometer, and magnetic sensor on a smartphone to improve positioning accuracy in urban environments. The main objective of the proposed method is to reduce the GNSS position error distribution, which is biased in the direction pointing across the street because of signal blocking in urban canyons. First, we calculate a positioning solution and its distribution using normal GNSS point positioning with a strict signal strength threshold and high-elevation mask angles. In GNSS positioning results, the horizontal position error distribution typically extends across the street. Multiple position candidates near the true position are then generated from the GNSS position error distribution. Next, we capture a photo using a smartphone that is aimed at the sky. The likelihood of the position candidates are then determined based on an image-matching technique that uses the captured photo and a virtual photo, which is simulated based on the candidate position, camera attitude, and 3D map. We use Google Earth to generate the virtual photo. Finally, the position candidate that has the highest likelihood is chosen as the final estimated user position. Using this technique, the error distribution of GNSS positioning is corrected by the camera, accelerometer, magnetic sensor, and 3D map. Thus, a more accurate user position can be determined. To confirm the effectiveness of the proposed technique, a positioning experiment was performed in a real-world urban-canyon environment. We used a GNSS receiver and a smartphone to collect data from multiple locations and evaluated the accuracy of the proposed technique. In this experiment, similar virtual photos can be generated at the true position from a 3D map. In other words, the likelihood of a position candidate increases when it is near the true position. In conclusion, our proposed method is effective for estimating accurate user positions in urban canyons, where conventional GNSS has large positioning errors.

AB - This paper describes a positioning technique that can be used for identifying pedestrians in urban environments in which there are large global navigation satellite system (GNSS) positioning errors because of poor satellite geometry and multipath signals. We propose a technique to realize position estimation based on GNSS and photos taken with a smartphone. We combine GNSS with a 3D map, camera, accelerometer, and magnetic sensor on a smartphone to improve positioning accuracy in urban environments. The main objective of the proposed method is to reduce the GNSS position error distribution, which is biased in the direction pointing across the street because of signal blocking in urban canyons. First, we calculate a positioning solution and its distribution using normal GNSS point positioning with a strict signal strength threshold and high-elevation mask angles. In GNSS positioning results, the horizontal position error distribution typically extends across the street. Multiple position candidates near the true position are then generated from the GNSS position error distribution. Next, we capture a photo using a smartphone that is aimed at the sky. The likelihood of the position candidates are then determined based on an image-matching technique that uses the captured photo and a virtual photo, which is simulated based on the candidate position, camera attitude, and 3D map. We use Google Earth to generate the virtual photo. Finally, the position candidate that has the highest likelihood is chosen as the final estimated user position. Using this technique, the error distribution of GNSS positioning is corrected by the camera, accelerometer, magnetic sensor, and 3D map. Thus, a more accurate user position can be determined. To confirm the effectiveness of the proposed technique, a positioning experiment was performed in a real-world urban-canyon environment. We used a GNSS receiver and a smartphone to collect data from multiple locations and evaluated the accuracy of the proposed technique. In this experiment, similar virtual photos can be generated at the true position from a 3D map. In other words, the likelihood of a position candidate increases when it is near the true position. In conclusion, our proposed method is effective for estimating accurate user positions in urban canyons, where conventional GNSS has large positioning errors.

UR - http://www.scopus.com/inward/record.url?scp=84975796976&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84975796976&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84975796976

VL - 4

SP - 2470

EP - 2480

BT - 28th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS 2015

PB - Institute of Navigation

ER -