COSBOS: COlor-Sensor-Based Occupancy Sensing

Introduction



The slides can be downloaded HERE (animations disabled in PDF file)

Publications

Quan Wang, Xinchi Zhang, Kim L. Boyer, Journal of Solid State Lighting 1:17, 2014. 
doi: 10.1186/s40539-014-0017-2

Quan Wang, Xinchi Zhang, Kim L. Boyer, 10th IEEE Workshop on Perception Beyond the Visible Spectrum (PBVS), 2014. 
doi: 10.1109/CVPRW.2014.46
(ORAL + Poster)

Quan Wang, Xinchi Zhang, Meng Wang, Kim L. Boyer, 22nd International Conference on Pattern Recognition (ICPR), 2014. 
doi: 10.1109/ICPR.2014.347
(ORAL)

Xinchi Zhang, Quan Wang, Kim L. Boyer, SPIE Optical Engineering + Applications, 2014. 
doi: 10.1117/12.2062105
(ORAL)



People

(Graduate Student)
(Undergraduate Student)
(Professor)

 
(Graduate Student)

 

FAQs

What is COSBOS?

COSBOS is the abbreviation of our COlor-Sensor-Based Occupancy Sensing technique. With multiple color-controllable LED fixtures and non-imaging color sensors, our technique enables low-cost and privacy-preserving occupancy distribution estimation. The direct application of this technique is occupancy-sensitive smart lighting, in which the system automatically delivers the light that best suits the occupancy scenario in an indoor space. This is a smart lighting system that can "think" - that can deliver the right light at the right time and the right place. 

What is perturbation-modulated lighting?

We add very small perturbations onto the visible light, such that these perturbations are imperceptible by human, but can be measured by color sensors. The change in color sensor readings under different perturbations is the key for estimating the occupancy. 



Why do we use color sensors?

Please take a look at this table:



What fixtures do we use?

We use the 7'' LED Downlight Round RGB (Vivia 7DR3-RGB) products from Renaissance Lighting. 

These fixtures are good for proof of concept and validation of methods. But they are not that fast. Using faster LEDs will let you develop more powerful real-time systems. 

What color sensors do we use?

For experiments in the JSSL paper, the PBVS paper and the ICPR paper, we were using SeaChanger wireless Colorbug sensors. However, these sensors are commercial products. They are expensive, slow, and not customizable. 

We have built our own color sensors based on Flora TCS34725 chips. We wire these chips to Raspberry Pi machines to make measurements. We are calling our own color sensors the "RPi sensors" internally. We use these sensors for real-time demos and the SPIE paper. 

How do we estimate the occupancy?

We implemented three different approaches:
  1. The machine learning approach (ICPR paper); 
  2. The light blockage model approach (JSSL paper and PBVS paper); 
  3. The light reflection model approach (JSSL paper). 
We achieved very good performance with all of these approaches. However, each approach has its limitations and strength. 

I am interested. Which paper shall I read? 

The JSSL 2014 paper is the most comprehensive one, covering the light blockage model and the light reflection model. 

The PBVS 2014 paper focuses on perturbation-modulated lighting and the light blockage model approach. 

The ICPR 2014 paper focuses on the sparse recovery of matrix A and the machine learning approach. 

The SPIE 2014 paper focuses on applications using the new RPi color sensor that we have built. 

Does the ambient light affect your performance?

No. The ambient light is considered in our light transport model. Our perturbation-modulated lighting eliminates the affect of ambient light. 

Is there an intellectual property (IP) on this work? 

We filed an US Provisional Patent Application on this work. 
  • Application number: 62/014745
  • Filing date: June 20, 2014


Links


Smart Lighting ERC: http://smartlighting.rpi.edu

Department of ECSE: http://www.ecse.rpi.edu

Rensselaer Polytechnic Institute: http://www.rpi.edu