A method for detecting pedestrian height and distance based on monocular vision technology. (August 2022)

• Record Type: Journal Article
• Authors: Shi, Zhan; Xu, Ziming; Wang, Tongxi
• Is Part Of: Measurement. Volume 199(2022)
• Publication Date: 2022-08
• Subjects: Background differencing -- Contour detection -- Monocular camera -- Monocular vision -- Target detection -- Measurement; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.111418
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: The method for detecting pedestrian height and distance information based on monocular cameras are reviewed. The process of using the background difference method and the contour detection method to optimize the detection effect is explored. The method to simplify application deployment in the form of configuration parameters are discussed. The effect of this method on measuring other objects is tested. Abstract: With the rapid development of smart cities, pedestrian detection information has expanded from human faces to human body features such as height due to the needs of urban security and urban management. In this paper, we propose a parametric method to detect the height and distance of pedestrians based on monocular cameras. First, we use target detection to detect the position of pedestrians in the figure. Then, the pedestrian's position in the frame is calibrated using a combination of an improved block differencing method and global contour detection to obtain the actual contour of the pedestrian with the complex background removed. After that, we use the perspective principle combined with the reference information to obtain the measurement height and distance formulas. Finally, we calculate the height of the pedestrian and the distance from the camera after checking the calibration. The experimental results show that the distance measurement accuracy is above 97%, and the height is above 95% a range of 25 m. The application of this method can evolve face detection to human body detection, extend the feature capture of the human face to the capture of human body height and other features, and improve the accuracy of face detection through the assistance of these features.