Download presentation
Presentation is loading. Please wait.
Published byPrudence Riley Modified over 5 years ago
1
Establishing an image-based ground truth for validation of sensor data-based room occupancy detection Steffen Petersen, Theis Heidmann Pedersen, Kasper Ubbe Nielsen, Michael Dahl Knudsen Published in 2016 Energy and Buildings Presented by Xiaoyu (Veronica) Liang
2
Backgrounds Buildings consume almost 40% of the energy used in most countries around the world. Heating, Ventilation, and Air Conditions (HVAC) systems are responsible for almost half of the energy consumed in buildings. Occupant based HVAC control approach can significantly improve energy savings. EuroStat 2010 Consumption by End Use EIA’s 2009 Residential Energy Consumption
3
Motivations Data-based occupancy detection methods need to be evaluated Automate the recording of the ground truth
4
Detecting persons in the cameral field of view
Proposed Method Detection Detecting persons in the cameral field of view Tracking Determines whether detected persons enter or leave the room
5
Detection Remove zero-value pixels
6
Detection
7
Tracking Kalman filter and munkres assignment algorithm
walls Camera field of view Kalman filter and munkres assignment algorithm A simple line crossing logic. Any detected moving object has to cross the Primary Virtual Lines and a Secondary Virtual Lines to be counted as entering or leaving the room. Robust against any small movement of objects in the image background. Primary virtual lines Secondary virtual lines walls
8
Performance Tests and Results
64-bit PC Intel I7-4710HQ CPU 16 GB RAM 30 frames per second Stress Test Determine the appropriate parameters. Function Test Test the ability of the proposed methods. Room Occupancy Test Test automatic real-time counting of occupants in normal office or dwelling rooms
9
Performance Tests and Results – Stress Test
Many persons crossing the camera field of vie within a short time period. Mount the camera at the top of main internal stair in a building with a high flow of persons. Results are compared with two independent and parallel series of manual people counting. Repeat multiple times to determine the best values. Results showed: 75 out of 76, entering 51 out of 52, leaving No False Positive & Not False Negative Accuracy 98.4%
10
Performance Tests and Results – Function Test
Detect and track one and multiple persons moving in and out of the camera field of view in different ways.
11
Performance Tests and Results – Room Occupancy Test
A 3-week test period. An office room occupied by 3 persons. Detect 296 entries and 297 exits. Compare with the parallel supervised observations (299 entries and exits) : 99.2% accuracy No False Positive No False Negative A 7-hours test period. An office room occupied by multiple persons. Detect 53 entries and 53 exits. Compare with the parallel supervised observations: 100% accuracy No False Positive No False Negative
12
Questions? Questions?
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.