The effect of pedestrian lighting on facial expression recognition with 3D models: A lab experiment. (15th January 2023)

• Record Type: Journal Article
• Authors: Li, Tianyu; Su, Hang; Zhang, Songbo; Xie, Binglei; Yang, Biao
• Is Part Of: Building and environment. Volume 228(2023)
• Publication Date: 2023-01-15
• Subjects: Pedestrian lighting -- Facial expression recognition -- 3D head model -- Spectral power distribution; Buildings -- Environmental engineering -- Periodicals; Building -- Research -- Periodicals; Constructions -- Technique de l'environnement -- Périodiques; Electronic journals; 696
• DOI: 10.1016/j.buildenv.2022.109896
• Languages: English
• ISSNs: 0360-1323

Abstract

Abstract: One function of road lighting is to enable pedestrians to evaluate the intentions of other people or to make interpersonal judgement. This paper aims to investigate the effect of lighting conditions on the ability of pedestrian to recognize facial expressions of 3D models. Fifteen lighting conditions were examined: five vertical illuminances (0.33, 1.00, 3.33, 10.0, and 30.0 lx) crossing spectral power distribution with three different scotopic/photopic ratios (0.57, 1.77 and 25.13). The results show that recognition of 3D facial expressions tended to be improved as illuminance increased. However, manipulation on spectral power distribution would not lead to better performance on 3D facial expression recognition. Compared to previous studies that used 2D images as visual stimuli, the current results from the 3D model present a slightly better performance on facial expression recognition. These findings are expected to contribute to the recommendations of optimal lighting levels based on facial expression recognition, and provide further evidence for the development of lighting standards. Highlights: Effect of pedestrian lighting on facial expression recognition with 3D models was investigated. Fifteen lighting conditions were examined: five vertical illuminances crossing three spectral power distributions. The 3D model can facilitate facial expression recognition than 2D photographs. Manipulation on spectral power distribution would not lead to better performance on 3D facial expression recognition.