Share

Enabling human-centered daylighting operation using non-intrusive luminance monitoring and deep learning.

Number: 3572

Author(s) : LU S., MAH D., TZEMPELIKOS A.

Summary

Luminance monitoring within the occupants' field of view (FOV) is required for assessing visual comfort and overall visual preferences, but it is practically challenging and intrusive. As a result, real-time, human-centered daylighting operation remains a challenge. No studies have determined if it is possible to acquire essential information on how conditions affect preferred luminance distributions, using a camera sensor placed in non-intrusive positions. This paper presents a novel deep-learning based framework method to demonstrate that meaningful features in the visual field can be extracted without invasive measurements or 3-D reconstruction of the occupant FOV. A Conditional Generative Adversarial Network (CGAN), pix2pix is used to transfer information from non-intrusive images to FOV images. Pix2pix takes a condition image (measured from a non-intrusive camera position) and generates a FOV image that is similar to a target image (measured from FOV). Two datasets were collected in an open-plan office with low-cost HDRI cameras installed at two alternate locations (a wall or a monitor), to separately train two pix2pix models with the same target FOV images. The results show that the generated FOV images closely resemble the measured FOV images in terms of pixelwise luminance errors, mean luminance, and structural similarity. The major source of error comes from some bright scenes visible through windows that are absent from non-intrusive images but appear in the FOV. However, more than 85% of the cases in both the training and validation set have low absolute luminance differences, given the inherently high window luminance under sunny conditions. This study is the first proof of concept demonstrating that it is possible to evaluate of visual preferences and enable human-centered daylighting operation without intrusive luminance monitoring, by employing the full potential of HDRI and deep learning techniques.

Available documents

Format PDF

Pages: 10 p.

Available

Free

Download

Details

  • Original title: Enabling human-centered daylighting operation using non-intrusive luminance monitoring and deep learning.
  • Record ID : 30032909
  • Languages: English
  • Subject: Technology
  • Source: 2024 Purdue Conferences. 8th International High Performance Buildings Conference at Purdue.
  • Publication date: 2024/07/15

Links


See other articles from the proceedings (63)
See the conference proceedings

Indexing