IEEE Spectrum shows what might develop into a mass market for 3D cameras - smart sliding doors, presented by University of Electro-Communications, and Hokuyo Automatic Co. at 2014 IEEE International Conference on Robotics and Automation (ICRA) in Hong Kong. A Youtube video explains the idea:
10 years ago this application was on the top 20 list for Canesta TOF sensors. Not sure how this is a new application. The problem is that automatic doors already exist and they work pretty well. It is not really a compelling app for TOF and the market potential is also low.
ReplyDeleteThe ToD camera has been commercialized by IEE for this kind of applications, the price is higher than door itslef ! Honestly, we can make such device with the smallest Log sensor associated to a simple uC and the cost will be a couple de dollars.
ReplyDelete-yang ni
Camera based door opening systems offer compelling values over the radar based systems.
ReplyDeleteBesides opening a door for a person, it counts the number of entrances/exits, detects piggybacks
and distinguishes people from animals/carts etc. It can also cross validate with a badge entrance and keep a copy of the badge holder face for example.
However it is hard to make it reliably due to variations in the lighting conditions especially when used for an exterior door when one side is illuminated by the sunshine.
GaryZ
Interesting features are not the same as compelling values. Please explain more what makes this compelling that automatic door manufacturers and their customers will pay the same or more for these extra features? Or, who makes these camera-based doors today? Certainly the technology to make a camera-based door (which I guess could be ToF or something else) has been around quite a while. These are serious questions and I'd like to know what you mean by compelling.
DeleteIf you look at the number of companies working on camera based surveillance for access control and asset tracking , you will appreciate the values of an integrated access control system.
DeleteJust take a look at 3VR.com for example.
A blind door opener for Wal-mart offers convenience, but loss prevention is paid separately with guards second-guessing.
OK, I am looking at this from an image sensor market point of view. Automatic doors are not going to be a killer app for ToF sensors, and that app is not new anyway. If you are in the security camera business along with integrated security systems, door control is of course and important application because people are paying a lot for security. Volumes can be low and market still fine due to high margins. Completely different conversation than where this blog post started.
DeleteIf you use a logarithmic sensor, all the lighting problems will go away.
ReplyDelete-yang ni
Linlog helps with overexposure, but the key issue seems due to the spectrum from the sun light.
ReplyDeleteWhen we have the iris based identification system in ATMs, we had very hard time to make it work
in bank lobbies with skylight.
Kinect with Primesense sensor also fail to operate reliably outdoors for the similar reason.
You are right, LinLog can only resolve the over exposition issue. For image analysis, a pure logarithmic response is needed. Since the focal plan image is formed by the reflection of amibient light S=R*L, if an image has pure logarithmic response then you can separate R & L terms. L is globally stable (in short term) and uniform, so the image content is mainly dominated by R which is invariant to illumination. That is why a pure log response is important in image analysis applications. That is also why all the biological vision systems use log response photoreceptors too. LinLog can not bring any advantage in such case, the interest of over-saturation alone is limited. We provide sensors with pure logarithmic response.
ReplyDelete-yang ni
Do biological receptors have log response or the Michaelis-Menton response?
Delete"That is also why all the biological vision systems use log response photoreceptors too"
ReplyDeleteThis is a pretty presumptuous statement. We can only guess why things evolved the way they did, and the reasons may be far more complicated, or more simple, than this reason. Perhaps instead you are a believer in intelligent design, or just biblical stories, in which case you cannot presume to know the mind of God unless he told you directly....
Intelligent design made by God should be similar to man-made things. That means the most intelligent design needs not to be a complex one and the smartness comes often from some hazardous combination which can be explained rationally afterword. One thing is sure that no biological vision uses linear photosensor, at least to our actual knowledge.
DeleteOur splendid color vision shows that non-linear sensor can give precise and vivid color sensing. We have reproduced such behavior by using our pure logarithmic sensor and this just gives me some personal conviction on this point.
Of course no one can tell what happens exactly in God's design as David Marr stated in his book Vision. He stated that one can understand an image processing algorithm using FFT running on complex CPU by just probing all the signals on the accessible nodes...
-yang ni
"He stated that one can understand an image processing algorithm using FFT running on complex CPU by just probing all the signals on the accessible nodes..."
ReplyDeleteit should be "... can NOT understand ..."
Sorry for this mistake !
Eric, why do you think that this assumption is presumptuous please ? It has been stated by a lot of biologists including David Hubel a Nobel winner. Please refer to :
http://hubel.med.harvard.edu/index.html
-yang ni
ha, I don't know Hubel but if he also said the same thing then I think he is being equally presumptuous...or even more so. Being a Nobel prize winner does not make you omniscient. Understanding "what" and "how" are quite different than understanding "why". Mankind has long presumed to answer "why" only to find that these explanations fall short in light of greater knowledge.
DeleteYou can say that log response in photoreceptors is efficient and useful. But the why is probably due to some random mutation when we all floated in the ocean. The efficiency and usefulness of that mutation gave some advantage over existing light sensing perhaps, and perhaps other mutations that came later did not give a compelling survival advantage. I don't presume to know why either. It just is.
Also, there are many meanings of the word presumptuous. In this case I meant "you presume to know too much", esp. of things that I believe are unknowable at this time.
I certainly do not understand much of what I know. Thinking otherwise would be presumptuous.
ReplyDeletePlease DO continue... that was such an AWESOME chat. :-)
ReplyDeletewhat is awesome here?
DeleteTwo smart people chatting is always awesome.... especially in such a specific field of which I want to learn so munch!
ReplyDelete