A METHOD FOR MEASUREMENT OF TRANSIENT DISCOMFORT GLARE CONDITIONS AND OCCUPANT SHADE CONTROL BEHAVIOR IN THE FIELD USING LOW-COST CCD CAMERAS

Transcription

1 A METHOD FOR MEASUREMENT OF TRANSIENT DISCOMFORT GLARE CONDITIONS AND OCCUPANT SHADE CONTROL BEHAVIOR IN THE FIELD USING LOW-COST CCD CAMERAS Kyle Konis, Ph.D Portland State University 1914 SW Park Ave Shattuck Hall Ste 207c Portland, Oregon ABSTRACT The transmission of sufficient daylight without creating visual discomfort is the central objective of effective daylighting in buildings. As emphasis on the performance of new and existing buildings increases, it is important to have accessible field-based methods for measurement and evaluation of discomfort glare in buildings. The use of High Dynamic Range imaging (HDR) has significant potential to improve the level of daylighting and discomfort glare measurement and analysis in the field. However, existing methods for HDR acquisition are often costly, and require connection to a PC for time-series image acquisition (a key capability when measuring transient daylighting conditions). This paper presents a method for autonomous acquisition of HDR images over multiple months using lowcost consumer digital CCD cameras without the need for additional hardware. The method is demonstrated using HDR data from a field study of a open-plan sidelit office building, where luminance maps acquired at 5-minute intervals are analyzed to assess visual comfort performance in response to transient daylighting conditions. This method can be utilized by students, educators, and practitioners to scale up HDR imaging from a research tool to a practical tool for daylighting performance monitoring in the field. 1. INTRODUCTION Assessing daylighting performance encompasses a range of factors that, if considered in isolation, can lead to misleading conclusions. For example, a design that maximizes daylight transmission to reduce electrical lighting energy consumption but results in visual discomfort for occupants may lead to constant use of interior shading devices as well as ad hoc modifications to the facade (or workstations), limiting both daylight availability and visual connection to the outdoors. Although there is a growing consensus for the need to include glare assessment in the evaluation of daylit commercial buildings, there remains a lack of practical tools to measure discomfort glare in the field. In previous studies, measurements of global vertical illuminance have been found to correlate with subjective assessments of discomfort glare (1). Additionally, measurement of surface luminance contrasts using a luminance meter is recommended as a method to evaluate visual discomfort in the field (2). Finally, a shielded illuminance sensor is recommended to measure the average luminance of largearea glare sources (3). However, these methods are often impractical in environments with transient daylighting conditions, where potential glare sources are multiple and constantly changing in size and location. High Dynamic Range (HDR) images store luminance data on a per-pixel scale, enabling both the definition and analysis of an arbitrary number of pre-defined regions within the camera s field of view, and the possibility of detecting glare sources to compute discomfort glare metrics. In computer graphics, the original HDR format (Radiance RGBE) was developed for the lighting simulation engine Radiance in order to record the photometric conditions of synthetic lighting environments (4,5). In following years, techniques were developed to produce HDR images from real-world scenes (6,7) by compositing multiple, exposurebracketed, low dynamic range (LDR) images (e.g., JPEG) into a single HDR image. Motivated by the possibility of using HDR in real spaces for photometric analysis, methods are now available to produce calibrated HDR images, commonly referred to as luminance maps (8).

2 The method is demonstrated using HDR data from a field study of an open-plan sidelit office building, where quantitative data from luminance maps acquired at 5-minute intervals are compared to subject assessments from building occupants to assess visual comfort performance. A detailed description of the procedure used to pair HDR data with occupant subjective data is described by Konis (12). The following sections discuss the method for time-series acquisition of HDR images. 2.1 Digital Camera Resulting from the need to monitor the workstations of 15 study participants simultaneously, a relatively low cost digital camera and lens converter were selected. The digital camera used in this study is the Canon PowerShot A570. This is a consumer-level camera that can be obtained for less than $100. To more-closely represent the human field of vision, a relatively inexpensive ($50) wide-angle lens converter (Opteka HD² 0.20X Professional Super AF fisheye lens, real angle of view = 174 deg.) was attached to each camera, and the camera/lens assembly was installed inside a mountable project enclosure as shown in fig. 2. A sample view from an occupant workstation acquired during the study is provided in fig. 3. Fig. 1: LBNL shade commissioning cart. Image from (9). Several efforts have been made to implement HDR imaging as a tool for glare analysis in the field. In 2007, a method was developed by the Lawrence Berkeley National Laboratory to commission and verify the performance of automated roller shade systems in the field using a mobile cart (fig. 1) with a digital camera, fisheye lens, and laptop computer (9). In 2008, Painter et. al., developed a method for evaluation of visual comfort in office workstations by mounting a digital camera adjacent to each participant s workstation and using a Mac mini to automate image acquisition in coordination with an on-screen glare survey (10, 11). The primarily limitation of existing approaches to HDR acquisition in the field is the cost of the computer needed to automate image acquisition. The requirement for one PC associated with each camera makes human factors studies involving multiple study participants impractical for many students and researchers. Fig. 2: HDR-enabled autonomous digital camera and fisheye lens converter shown inside project enclosure with mounting bracket. Note: the small white box is a temperature/rh sensor taped to the mounting bracket. 2. METHOD This paper presents a method for acquisition of HDR images over multiple months using low-cost consumer digital CCD cameras without the need for an associated PC. 2

3 using a Radiance process to calculate the average luminance of the identical window region as viewed from the camera (fig. 4). Fig. 3: Example view from participant workstation located in the core. Luminance values are represented with a falsecolor log-scale where yellow indicates values above 4000 cd/m Camera Control To enable acquisition of HDR images without external PC control, the Canon camera firmware was modified using a subset of the Canon Hacker s Development Kit (CHDK) called Stereo Data Maker (SDM). SDM is a firmware enhancement that is loaded into the camera s memory upon bootup and provides additional functionality beyond that currently provided by the native camera firmware (13). Using a scripting language based on Adam Dunkel s ubasic (14), SDM users can precisely control a wide range of camera features automatically. In this study, a custom script was written to automate acquisition of exposure-bracketed sets of JPEG images at regular (e.g. 5 minute) intervals and to automatically wake up and put to sleep the camera based on a user-defined schedule. Bracketed sets of JPEG images were stored on the camera s 8GB SD memory card. Based on the image size and quality settings chosen, and a daily acquisition interval from 6AM to 6PM, (10 JPEGs every 5 minutes), each camera was capable of operating for over one month autonomously. At the end of each study interval, bracketed sets of JPEGs were downloaded to a PC and composited into HDR images using the Radiance lighting simulation software program hdrgen. Fig. 4: Shielded illuminance sensor (right) and masked region used to calculate the average window luminance for comparison to the shielded sensor (left). The average luminance of the window region calculated from each HDR image was then compared with a reading acquired simultaneously from the reference sensor. A comparison is illustrated in fig. 5. The results show that when the average window luminance exceeds 500 cd/m 2, average window luminance calculated from the HDR image is, on average, 18% below the value recorded by the reference sensor. After a global scaling of each HDR image by a coefficient of 1.22, the HDR image data are comparable to the reference (mean = 0.1%, SD = 11%, max = 44%, N = 88 images). However, when the target is below 500 cd/m 2, the accuracy of the HDR images diminishes significantly, and consistently overestimates scene luminances (mean = 40%, SD = 23%, max = 66%, N = 25 images). Because the analysis of HDR data for discomfort glare typically focuses on bright lighting conditions, where the average luminance of the potential glare source is likely to be significantly greater than 500 cd/m 2, the HDR data are considered to be of acceptable level of accuracy and measures are expected to be within + / - 10%. 2.3 Calibration Prior to the field study, a study was conducted over a day with clear sky conditions from 8:00 to 17:00 PST in one cell of the Lawrence Berkeley Windows and Daylighting Testbed facility to quantify measurement errors associated with use of these HDR data. A procedure was developed to compare the HDR data derived from the digital cameras to a calibrated shielded illuminance sensor. The shielded illuminance sensor was masked to measure the average luminance of the window region of the test cell (excluding the lower 30 inches). A camera was located adjacent to the shielded illuminance sensor and a mask was created 3

4 Region Luminance (cd/m^2) Percent Error Luminance Data from A570 vs. Shielded Illuminance Sensor N = 107 Fig. 5: Comparison of measures of average window luminance from HDR image data to reference shielded illuminance sensor before and after HDR data were uniformly scaled by a calibration coefficient (coef. = 1.22). 3. SAMPLE RESULTS The following section presents sample results that demonstrate the applicability of this method for visual comfort analysis in the field. In addition, because luminance maps also record contextual data, the ability to composite sets of time-series HDR images into movies enables visual inspection of behavioral modifications to the indoor environment over time. Therefore, time-series HDR acquisition also presents a practical method for measurement of occupant operation of windows and interior shading devices. 3.1 Visual Discomfort Hour Measured Luminance Calibrated A570 Luminance Orig. A570 Luminance Percent Error (before calibration) Percent Error (after calibration [coef. = 1.22]) Hour of the upper daylight zone windows shown in fig. 3 to three discomfort probability thresholds (p = 0.2, p = 0.5, p = 0.75). The discomfort thresholds were determined from logistic regression models of subjective response data paired with luminance data. A detailed description for how the logistic models were developed is provided in Konis (12). Luminance (cd/m2) (p = 0.75) (p = 0.5) (p = 0.2) Fig. 6: Maximum measured upper window luminance for a day with predominantly clear sky conditions, showing discomfort probability thresholds (p = 0.2, p = 0.5, p = 0.75). 3.2 Occupant Behavior Maximum upper window luminance Discomfort thresholds In addition to visual comfort performance, time-series HDR imaging in the field enables behavioral modifications to the building facade to be examined visually and without the loss of visual information common in low dynamic range images of scenes with high visual contrast. For example in this study, images were used as a practical way to record numerous interior roller shade control events in response to changes in outdoor weather conditions over multiple days. Fig. 7 presents a condensed visualization of time-series images over multiple days. Fig. 8, represented with a falsecolor luminance scale, illustrates one instance of occupant shade control. These events can be paired with physical measures to examine and model occupant control of shading devices in response to interior and exterior environmental stimuli. The monitoring method discussed in this paper has also been applied to record the window-openfraction of multiple operable windows in a naturally ventilated office building to examine the relationship between observed occupant control behavior and existing window behavioral models (15). Hour To investigate the relationship between quantitative indicators of visual discomfort and occupant subjective assessments, HDR images from each occupant workstation were analyzed to compute a range of emerging and existing quantitative indicators of visual discomfort. As one example, fig. 6 compares the maximum window luminance 4

5 example of the occlusion index and fig. 10 shows summary results for a number of observed facade sections. Fig. 7: Condensed visualization of time-lapse imaging from a viewpoint of one facade section (showing one image per hour rather than one image every 5 minutes). Fig. 9: The facade occlusion index developed to summarize patterns of operation of shading devices based on visual inspection of time-series HDR images. Numbers shown in green indicate the number of times each individual shade was operated during the monitoring interval. The number in brackets [72%] indicates the average facade section occlusion index. Fig. 8: Time-lapse movies of shade control were observed and changes to shade position logged in a spreadsheet format. 3.3 Facade Occlusion To estimate the actual levels of visible light transmittance enabled by a glazed facade, time-series HDR imaging can be used to systematically record the position of interior shading devices. To summarize observed patterns of shade use, an occlusion index was used to code the shade position systematically from visual inspection of each set of HDR images. The term occlusion in this context refers to the blocking of a glazed opening by an interior roller shade. The occlusion index denotes the percent area of each individual window section that is occluded by roller shade fabric or furniture at a given point in time on a scale of 0 to 100% (where 100% denotes full occlusion, i.e. the shade is fully deployed to cover the window). Fig. 9 presents an Fig. 10: Occlusion index visualization for a portion of facade sections observed over a three-week monitoring interval. 4. DISCUSSION The use of daylight to reduce energy consumption and enhance Indoor Environmental Quality (IEQ) is one of the most common claims made for commercial office buildings promoted as sustainable, energy efficient, green, or high performance. Rarely are daylit buildings examined in use to compare design intent to measured performance. The method presented in this paper extends the capacity for performance monitoring of daylit buildings by demonstrating a low-cost method for time-series acquisition of HDR images. While visual inspection of HDR images presents a useful means for recording occupant behavior in response to transient environmental conditions, the process is time-consuming and could be improved through the application of image processing algorithms capable of detecting the adjustment of shading devices or operable windows. This is an area for future research. 5

6 Although the method described in this paper enables timeseries acquisition of HDR data, these data are only as useful to designers as the visual discomfort models or criteria used to interpret them. There remains limited consensus for what discomfort models are appropriate, as well as how quantitative data from discomfort models should be interpreted in daylit environments where occupants are exposed to a range of stimulus conditions over the course of a single day. The method presented in this paper demonstrates one possible tool to validate and improve existing assumptions for visual comfort in daylit work environments. 5. CONCLUSIONS The ability to acquire time-series HDR images from lowcost consumer digital cameras without a PC significantly extends the capacity of students, researchers, and professionals to conduct performance monitoring of daylit buildings in use. Although time-series acquisition of HDR images in real work environments enables a greater level of lighting data to be examined, effective analysis is reliant on validated quantitative criteria and models of visual discomfort. 6. ACKNOWLEDGEMENTS Funding for the cameras and associated equipment for this project was provided by a grant from the T.Y. Lin foundation. Support for fabricating the project enclosures was provided by Howdy Goudey of the Lawrence Berkeley National Laboratory. 7. REFERENCES (1) Osterhaus, W.K.E., & Bailey, I.L., Large area glare sources and their effect on visual discomfort and visual performance at computer workstations. Proceedings of the IEEE Industry, 1992 (2) Rea M, editor, The IESNA Lighting Handbook. 9th ed. New York: The Illuminating Engineering Society of North America, 2000 (3) IEA SHC Task 21, Daylight in buildings, Appendix, (4) Ward G. Radiance Visual Comfort Calculation. < Accessed 10/02/2011 (5) Ward G., The RADIANCE Lighting Simulation and rendering system. Computer Graphics, Proceedings of Siggraph, 1994 (6) Debevec PE, Malik J.,Recovering high dynamic range radiance maps from photographs. Proceedings of Siggraph, 1997 (7) Mitsunaga T, Nayar SK, IEEE Conference on computer vision and pattern recognition (CVPR), Vol.1, , 1999 (8) Inanici M, Galvin J. 2004, Evaluation of high dynamic range photography as a luminance mapping Technique. Lawrence Berkeley National Laboratory. < Accessed 01/10/2011 (9) LBNL, Commissioning and Verification Procedures for the Automated Roller Shade System at The New York Times Headquarters, New York, New York, May 30, 2007 (10) Painter, B., Fan, D. and Mardajevic, J., Minimally intrusive evaluation of visual discomfort in the normal workplace, 7 th international Radiance workshop, Fribourg, Switzerland, 2008 (11) Fan D, Mardaljevic J. 2009, HDRcapOSX. How to automate capture using a Mac Mini less k/b and screen. < rcaposx> Accessed 3/14/2012 (12) Konis, K.S., Effective Daylighting:Evaluating Daylighting Performance in the San Francisco Federal Building from the Perspective of Building Occupants. Doctoral Dissertation, University of California, Berkeley. Published 2011 (13) Stereo Data Maker (SDM) < Accessed 03/15/2012 (14) ubasic < Accessed 03/15/2012 (15) Dutton, S., Smires, Y., Konis, K., Zhai, S., Zhang, H., Haves, P., Application of a Stochastic Window Use Model in EnergyPlus. Submitted to Simbuild