Cascaded CNN method for far object detection in outdoor surveillance

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

Abstract

In maritime surveillance, detection of small ships and vessels located far away in the scene is of vital importance for behaviour analysis. Comparing to closely located objects, far objects are often captured in a smaller size and lack the adequate amount of details. Therefore, conventional detectors fail to recognize them. This paper proposes a CNN-based cascaded method for reliable detection of objects and more specifically vessels, located far away from a surveillance camera. The cascaded method improves small object detection accuracy by additional processing of the obtained candidate regions in their original resolution. The additional processing includes another detection iteration and a sequence of detection verification steps. Experimental results on our real-world vessel evaluation dataset reveal that the cascaded method increases the recall rate and F1- measurement by 13% and 12%, respectively. Another benefit is that the method does not require an adopter to change the model and architecture of the applied network. As an additional contribution, we provide a labeled maritime dataset to open public access.

LanguageEnglish
Title of host publicationProceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018
EditorsRichard Chbeir, Gabriella Sanniti di Baja, Luigi Gallo, Kokou Yetongnon, Albert Dipanda, Modesto Castrillon-Santana
Place of PublicationPiscataway
PublisherInstitute of Electrical and Electronics Engineers
Pages40-47
Number of pages8
ISBN (Electronic)978-1-5386-9385-8
DOIs
StatePublished - 3 May 2019
Event14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018 - Las Palmas de Gran Canaria, Spain
Duration: 26 Nov 201829 Nov 2018

Conference

Conference14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018
CountrySpain
CityLas Palmas de Gran Canaria
Period26/11/1829/11/18

Fingerprint

Processing
Ships
Cameras
Detectors
Object detection

Keywords

  • Cascaded CNNs
  • Far object detection
  • Maritime surveillance
  • Vessel detection

Cite this

Ghahremani, A., Bondarev, E., & de With, P. H. N. (2019). Cascaded CNN method for far object detection in outdoor surveillance. In R. Chbeir, G. S. di Baja, L. Gallo, K. Yetongnon, A. Dipanda, & M. Castrillon-Santana (Eds.), Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018 (pp. 40-47). [8706233] Piscataway: Institute of Electrical and Electronics Engineers. DOI: 10.1109/SITIS.2018.00017
Ghahremani, Amir ; Bondarev, Egor ; de With, Peter H.N./ Cascaded CNN method for far object detection in outdoor surveillance. Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018. editor / Richard Chbeir ; Gabriella Sanniti di Baja ; Luigi Gallo ; Kokou Yetongnon ; Albert Dipanda ; Modesto Castrillon-Santana. Piscataway : Institute of Electrical and Electronics Engineers, 2019. pp. 40-47
@inproceedings{7687c60a380b41d3a2b37b81ce6a5d34,
title = "Cascaded CNN method for far object detection in outdoor surveillance",
abstract = "In maritime surveillance, detection of small ships and vessels located far away in the scene is of vital importance for behaviour analysis. Comparing to closely located objects, far objects are often captured in a smaller size and lack the adequate amount of details. Therefore, conventional detectors fail to recognize them. This paper proposes a CNN-based cascaded method for reliable detection of objects and more specifically vessels, located far away from a surveillance camera. The cascaded method improves small object detection accuracy by additional processing of the obtained candidate regions in their original resolution. The additional processing includes another detection iteration and a sequence of detection verification steps. Experimental results on our real-world vessel evaluation dataset reveal that the cascaded method increases the recall rate and F1- measurement by 13{\%} and 12{\%}, respectively. Another benefit is that the method does not require an adopter to change the model and architecture of the applied network. As an additional contribution, we provide a labeled maritime dataset to open public access.",
keywords = "Cascaded CNNs, Far object detection, Maritime surveillance, Vessel detection",
author = "Amir Ghahremani and Egor Bondarev and {de With}, {Peter H.N.}",
year = "2019",
month = "5",
day = "3",
doi = "10.1109/SITIS.2018.00017",
language = "English",
pages = "40--47",
editor = "Richard Chbeir and {di Baja}, {Gabriella Sanniti} and Luigi Gallo and Kokou Yetongnon and Albert Dipanda and Modesto Castrillon-Santana",
booktitle = "Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018",
publisher = "Institute of Electrical and Electronics Engineers",
address = "United States",

}

Ghahremani, A, Bondarev, E & de With, PHN 2019, Cascaded CNN method for far object detection in outdoor surveillance. in R Chbeir, GS di Baja, L Gallo, K Yetongnon, A Dipanda & M Castrillon-Santana (eds), Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018., 8706233, Institute of Electrical and Electronics Engineers, Piscataway, pp. 40-47, 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018, Las Palmas de Gran Canaria, Spain, 26/11/18. DOI: 10.1109/SITIS.2018.00017

Cascaded CNN method for far object detection in outdoor surveillance. / Ghahremani, Amir; Bondarev, Egor; de With, Peter H.N.

Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018. ed. / Richard Chbeir; Gabriella Sanniti di Baja; Luigi Gallo; Kokou Yetongnon; Albert Dipanda; Modesto Castrillon-Santana. Piscataway : Institute of Electrical and Electronics Engineers, 2019. p. 40-47 8706233.

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Cascaded CNN method for far object detection in outdoor surveillance

AU - Ghahremani,Amir

AU - Bondarev,Egor

AU - de With,Peter H.N.

PY - 2019/5/3

Y1 - 2019/5/3

N2 - In maritime surveillance, detection of small ships and vessels located far away in the scene is of vital importance for behaviour analysis. Comparing to closely located objects, far objects are often captured in a smaller size and lack the adequate amount of details. Therefore, conventional detectors fail to recognize them. This paper proposes a CNN-based cascaded method for reliable detection of objects and more specifically vessels, located far away from a surveillance camera. The cascaded method improves small object detection accuracy by additional processing of the obtained candidate regions in their original resolution. The additional processing includes another detection iteration and a sequence of detection verification steps. Experimental results on our real-world vessel evaluation dataset reveal that the cascaded method increases the recall rate and F1- measurement by 13% and 12%, respectively. Another benefit is that the method does not require an adopter to change the model and architecture of the applied network. As an additional contribution, we provide a labeled maritime dataset to open public access.

AB - In maritime surveillance, detection of small ships and vessels located far away in the scene is of vital importance for behaviour analysis. Comparing to closely located objects, far objects are often captured in a smaller size and lack the adequate amount of details. Therefore, conventional detectors fail to recognize them. This paper proposes a CNN-based cascaded method for reliable detection of objects and more specifically vessels, located far away from a surveillance camera. The cascaded method improves small object detection accuracy by additional processing of the obtained candidate regions in their original resolution. The additional processing includes another detection iteration and a sequence of detection verification steps. Experimental results on our real-world vessel evaluation dataset reveal that the cascaded method increases the recall rate and F1- measurement by 13% and 12%, respectively. Another benefit is that the method does not require an adopter to change the model and architecture of the applied network. As an additional contribution, we provide a labeled maritime dataset to open public access.

KW - Cascaded CNNs

KW - Far object detection

KW - Maritime surveillance

KW - Vessel detection

UR - http://www.scopus.com/inward/record.url?scp=85065925230&partnerID=8YFLogxK

U2 - 10.1109/SITIS.2018.00017

DO - 10.1109/SITIS.2018.00017

M3 - Conference contribution

SP - 40

EP - 47

BT - Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018

PB - Institute of Electrical and Electronics Engineers

CY - Piscataway

ER -

Ghahremani A, Bondarev E, de With PHN. Cascaded CNN method for far object detection in outdoor surveillance. In Chbeir R, di Baja GS, Gallo L, Yetongnon K, Dipanda A, Castrillon-Santana M, editors, Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018. Piscataway: Institute of Electrical and Electronics Engineers. 2019. p. 40-47. 8706233. Available from, DOI: 10.1109/SITIS.2018.00017