Towards a scalable resource-driven pedestrian detection for low-cost IR sensors

Jia Qu; Shotaro Miwa

doi:10.1117/12.2663441

13 June 2023 Towards a scalable resource-driven pedestrian detection for low-cost IR sensors

Jia Qu, Shotaro Miwa

Proceedings Volume 12534, Infrared Technology and Applications XLIX; 125341W (2023) https://doi.org/10.1117/12.2663441
Event: SPIE Defense + Commercial Sensing, 2023, Orlando, Florida, United States

Conference Poster

Abstract

Systems using low-cost infrared (IR) sensors have gained more and more interest in recent years, such as smart video surveillance, driver assistance system, robotics, and military applications. To meet the total cost limit of these systems, low resource processers are desired. In this paper, we propose a scalable resource-driven pedestrian detection approach for low-cost IR sensors. The proposed method introduces a feature-importance-sampling based feature selection for the IR sensors which enables scalable selection of the number of effective features according to the resource requirements. To obtain high performance detection with the few selected effective features and hard-to-obtain data of low-cost IR sensors, we adopt transfer learning from easy-to-obtain visible images to train the pedestrian detector. In this way, optimal IR system design can be determined based on the trade-offs between performance and resources. Using HOG and SVM based pedestrian detection methods, we evaluate the proposed method on Thermal Diode lnfraRed (TDIR) sensor and find that key-points of the human body that observed in the TDIR images, such as the contours of the body, are selected as the most crucial attributes. As a result of training these selected features on visible images, one can reduce memory use by 1/30 and increase detection speed by about 1/6 while maintaining competitive detection performance. To summarize, the proposed work presents a scalable resource-driven method for pedestrian detection for low-cost IR sensors with high accuracy and low resource consumption.

Citation Download Citation

Jia Qu and Shotaro Miwa "Towards a scalable resource-driven pedestrian detection for low-cost IR sensors", Proc. SPIE 12534, Infrared Technology and Applications XLIX, 125341W (13 June 2023); https://doi.org/10.1117/12.2663441

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available