Improving the visibility of nighttime images for pedestrian recognition using in‐vehicle camera. Issue 10 (31st August 2020)

• Record Type: Journal Article
• Authors: Ogura, Ryota; Nagasaki, Takeshi; Matsubara, Hitoshi
• Is Part Of: Electronics and communications in Japan. Volume 103:Issue 10(2020)
• Page Start: 35
• Page End: 43
• Publication Date: 2020-08-31
• Subjects: deep learning -- image conversion -- monocular camera -- pedestrian detection; Electronics -- Periodicals; Telecommunication -- Periodicals; 621.6805
• DOI: 10.1002/ecj.12268
• Languages: English
• ISSNs: 1942-9533

Abstract

Abstract: There is a need for methods to recognize night pedestrian to reduce pedestrian traffic accidents at night. In this paper, we proposed the method that converts images using continuous nighttime images from in‐vehicle camera. The proposed method performs feature extraction that does not depend on one's own vehicle speed because inputting continuous nighttime images, and the function of convolution layer that performs dimensionality reduction. In order to confirm the effectiveness of the proposed method, we prepared the images of simulation and camera. The nighttime images were conversed with the proposed method. After conversion, we calculated the recognition performance by applying object detection which is an object recognition method. We showed the proposed method is more robust against the own vehicle speed change.