Additionally,Mobotix recommends that subjects pause for 1-2 seconds for measurement and do not continue moving for best accuracy. When measuring subjects on the move, reported temperature dropped significantly.
We have developed software to work with the Mobotix cameras to enhance the solution. By using face detection and facial landmarks to focus on specific areas, we remove a lot of the extraneous readings. We have integrated black body systems and also created the option for a "virtual" black body whereby one can choose any area within the camera view to create a reference (in temperature stable environments). Also included in the solution are on-screen indicators, and "trigger" points for event notifications.
This solution definitely assists in overcoming some of the shortcomings listed in the article.
Perhaps, but extra software cannot really overcome the fact that temperature cameras have really turned out to be a farce in terms of mitigating COVID spread.
I suppose there will be debate as to whether ANY temperature analysis methods will prevent/limit the spread of Covid, so there will be differing opinions. I can't comment on this, only on the technology solution.
We have talked to other manufacturers that have done the same with Mobotix cameras, there are a couple concerns related to the screening methods Cathexis demonstrates; multiple person screening and on the move screening. In the demonstration video on the Cathexis COVID solution page, the tracking indicates forehead (person wearing glasses), obscured faces, mouth/hottest point scanning:
,
This does not match Mobotix recommendations for single subject stop-and-scan, however that video seems to be marketing a Dahua-based solution? Can you clarify this?
Also as our tests showed okay performance finding the inner canthus, can you give some examples of "we remove a lot of the extraneous readings"?
As you have pointed out, screening "on-the-move" is not the best solution. The stop-and-go solution definitely provides the the best results (as per the Mobotix recommendations).
It is very difficult to identify facial landmarks on the go, so in these scenario's we use only face detection and face size to initiate the point at which the reading is taken
We do have customers who use the measurement on the move, and they understand and accept that there is a reduction in accuracy. We use the face detection to try and improve results by ignoring readings until the face is at a certain size, so that we don't get a whole lot of incorrect logs while the person is walking towards the camera.
As you correctly spotted, the video shown is not from the Mobotix camera solution.
To answer your question regarding the extraneous readings. We receive readings from Mobotix for the complete camera view. We then use the analytics to "focus" our attention on specific areas, thereby removing miscellaneous items like hot coffee and readings from other areas of the person or camera view. In order to identify the facial landmarks, the subject needs to be pretty much paused in front of the camera.
Another thing that we had to take into consideration was the "parallax error" due to the two Mobotix lenses being slightly offset from one another. Because we are using the visible light camera for the analytics but receiving the temperature readings from the thermal camera we had to introduce parallax error compensation calibration.
We have a lot of deployments in the EU region with the solution, so if you need any more detail from the guys there who have real-world experience, I can put you in touch.
Regards
Gus
Agree
Disagree
Informative
Unhelpful
Funny
Create New Topic
Subscribe to IPVM Research to read the full report.
Why do I need to subscribe?
The IPVM Research Service includes products tests and shootouts plus competitive and financial analysis, helping decision-makers better evaluate purchasing, partnering, developing, and/or competing against companies in physical security.
Comments (6)
Guilherme Barandas
What sensor was used on the m16? Because the M16 has the Thermal R sensor that from what I read, dispenses the Black Body. Am I right?
Create New Topic
Gus Brecher
We have developed software to work with the Mobotix cameras to enhance the solution. By using face detection and facial landmarks to focus on specific areas, we remove a lot of the extraneous readings. We have integrated black body systems and also created the option for a "virtual" black body whereby one can choose any area within the camera view to create a reference (in temperature stable environments). Also included in the solution are on-screen indicators, and "trigger" points for event notifications.
This solution definitely assists in overcoming some of the shortcomings listed in the article.
Create New Topic