I received a request from reader Kirk B. to post a follow-up to the Q&A with LEGO regarding the question about the camera - LEGO pushed back and asked what WE (the consumers) were looking for...
Kirk writes:
The LEGO Co. had an interesting response to the LEGO camera question Jim Kelly asked them at the FLL World Festival:
3. Is there a plan to create a cheap (<100$) camera sensor for autonomous robot operation?I would turn the question around and ask: what would you really want to do with a ‘camera sensor’?
Reading between the lines, it seems the LEGO Co. is looking for a convincing argument(s) to create a new camera.
We could help the LEGO Co. by providing them some useful information for the development of a new LEGO camera. (Once all the comments are received, they will be forwarded to the LEGO Co.).
Please comment what improvements you would make to the original LEGO camera and its software, and how much you would be willing to pay for the new LEGO camera. Also, “what would you really want to do with a ‘camera sensor’?”
I agree with Kirk wanting to pursue this, so if you've got something to add regarding the camera, please do so here... Kirk and I will try to forward this information back to LEGO in 2-3 weeks and see what they say.
15 comments:
Just dreaming, I'd like to use a camera for wireless internet, remote surveillance. I realize that would probably reqire other hardware (like Charmlab's Qwerk controller, for example).
But wireless, remote surveilance from your NXT is the dream for me.
Wow, I can think of several awesome uses for a wireless camera if Lego made one - I'm thinking of a wireless camera that has Vision Command-type abilities.
- Motion detection: a camera would be able to detect and track motion much better than an Ultrasonic sensor. For example, an Ultrasonic sensor can't track an object moving around, but a camera could do it no problem. This would be really useful for making robots that could follow other robots or people, for example.
- Color sensing: camera would be very handy for detecting different colors in several different places.
- User-control features: a camera would open up WAY more possibilities for user-controlled robots, by giving the user a view of the surroundings.
- Combo Sensor: A camera can act as a light sensor, color sensor, motion sensor, and object sensor (?) all in one. This greatly increases the abilities of a robot, as it lets the robot have more sensors.
Is that good? I've been dreaming about a wireless camera like Vision Command since I first got the RIS - I would probably be willing to pay about $100 for one if Lego made it.
-Jonathan
I built a three wheeled robot that is operated remotely using a second Lego Intelligent Brick and Ultratechnic's Tilt Sensor. I want to be able to run the robot from place to place and be able to see where it is going on the Remote's screen. If Lego would produce the Camera Sensor that would do the above I would be one of the first to buy it. <$100 would be a good price.
I could use the camera sensor to make obstacle avoidance systems based on robot vision. Motion detection, face detection, control based on light sources. I could even think of processing the information gathered from the camera sensor and feeding it to a neural network or a cellular automata to do a lot of things.
Image processing, Machine Learning, Robot Cooperation... There are a lot of things that could be done!!!
Ronald Delgado.
Caracas, Venezuela.
So nobody has any good uses for a LEGO camera?
Remote surveillance and remote operation are easily done using any wireless camera. What is the added value from LEGO, easy mounting and power from the NXT?
Motion detection, color sensing and object recognition would be cool. But that will require the camera sensor to have a lot of memory and a powerful processor. Sounds spendy. How do you program the camera, and how would this information be conveyed to the NXT? What is the format and transport mechanism? What would a LEGO solution give you that you can't do right now with a wireless camera and vision processing software on your laptop?
Basically I want a CMUCam2...getting that to communicate w/ the NXT, and keeping it under $100 would be difficult.
So if that's not the option, here are the capabilities I want:
1. Can draw power from both Battery pack and NXT brick (motor port A...Thanks Philo!)
2. Can communicate with Brick in a closed loop. Sensor port 1.
3. Has ability to send information to a computer/video monitor.
4. Vision Command capabilities w/ on board processing.
5. I want it to match JP Brown's Rubic's Cube solver...without the computer...make it just the camera and the NXT brick.
ynI would like a camera for survelience--to build a security bot that can detect intrusion, follow and intruder, and wirelessly transmit pics to a receiver.
The open-source AVRCam is what I had in mind. All it does is on-the-fly blob recognition. This does NOT require a memory big enough for a single frame! Other algorithms which can be implmented w/o frame buffer are discrete cross-correlation (track camera motion) and probably more. The BOM of such device in mass-production, surface-mounted electronics and manufacturing in the east (as all Lego electronics..) would substaintaly less then 100USD. The advance of mobile camera technology can allow putting better, higher resolution camera chips at low cost. It would be nice to have video streaming (e.g. MPEG4 compressed) via RS485 port 4 and BT to a PC/Mac but not a must. Also object recognition will be great, maybe with something simpler like using algorithms from augmented reality open-source packages like ARTag.
To try to put the camera issue in perspective, though I may be incorrect, I can see 2 main options for LEGO to pursue if they decide to release a camera:
1. Have the camera with a built in processor for on-board image analysis to directly command the NXT to respond. (It may be useful to have the camera and processor separate so the camera can be replaced/upgraded.) Also the camera should have wireless communication to a computer for observation by the user.
2. Have the camera send wireless images to the computer for image analysis or observation by the user (no on-board processor for the camera on the robot).
I would be happy if LEGO would sell any of the above camera systems, but I prefer (1). I would be willing to pay up to $150 USD for (1), but only up to $100 for (2). My key point is that the camera has to be wireless! (This would probably require LEGO to offer some kind of wireless receiver as a separate purchase.)
More importantly, what LEGO would have to do is have a “plug and play” solution, as they would be marketing this to children, primarily. Not too many children could be able to program the vision processing softwares that are currently available, and get the vision program to command the robot what to do. (Many adults are looking for a “plug and play” solution, as well). This is what a LEGO SOLUTION COULD PROVIDE that no one else can currently provide.
Equally important, the vision system should be programmable in NXT-G with the inclusion of some basic image analysis blocks, such as for motion, object, and colour detection. Also, the Mindstorms community could create their own vision blocks (facial expression recognition, etc). National Instruments has a machine vision software package that is an add-on to LabVIEW, hence it makes sense for LEGO to have them produce the NXT-G machine vision add-on.
And if LEGO will not have a camera, I want to at least have an NXT-G vision software add-on that can work seamlessly with NXT-G, and be compatible with a variety of camera systems.
You know, there was a Mindstorms camera for use with the RIS that you could use as a sensor. I don't have it but I've heard it had built-in IR so the RCX could send msg's to it so it would start filming or take a pic that was sent via a USB cable to the PC. It had a software that made you able to program your bot to use the camera sensor as either a "distance" sensor like thing and as a color sensor. It has a BAD resolution in pixels and the quality of the pix and movies was poor. But ofcourse, rather that then nothing. I only hope it use BT instead of USB to send movies and pix to the PC so it could drive where ever you want it to as long as it stays in range. I also hope they make a add-on or a separate software that let's you take advantege of the camera sensor and let's you program the robot so it could be automnus. Just my cents.
Andy
I will also add:
There is one thing for certain, that LEGO would have to offer a better camera than the original camera release (Vision Command) before anyone would start buying it, or at least myself.
The new camera, at minimum, would have to be wireless (something with a higher transmission rate than BT, such as Wi-Fi), have a higher resolution, work in low light conditions, and have improved vision software.
Not necessary, but I would like it to have infrared LED's for night vision, and a built in microphone. The signal from the microphone should also be transmitted wirelessly to the computer (in surveillance purposes, its nice to not just see, but hear too).
To help keep the cost of the camera expansion kit down, no need to supply any Technic parts, as was done with Vision Command.
Powered via the NXT (using Philo's method would work).
Have it connect to a sensor port on the NXT? That depends on what I want the NXT to do with it. I don't think the NXT brick is powerful enough to handle much of the video processing.
I would like to see a package deal.
A LEGO form factor small color camera. A small USB device to receive signals from the camera and feed them into the PC (these are already available). And a good communication protocol to send commands and input back to the NXT wirelessly (bluetooth would work, it just needs to be simple to implement).
But, I would also like to see this capable of handling multiple camera signals.
I think that starts getting pricey.
I don't want the camera to send the signal back to a PC. I want it on a robot connected as a sensor to the Intelligent Brick which would transmit the picture to another Intelligent Brick used as a remote thorugh Bluetooth. That is what I want a camera sensor to be able to do. I don't want to have to carry a laptop around to see where the robot is going as I am remotely operating a robot. It would mean two way communication- remote Brick communicating angle of movement for the robot Brick and the robot Brick sending pictures back to the remote control Brick.
What I would really like from the camera function would be just to have it take a picture when the program states. Then store it on the NXT, and finally allows to take it off of the NXT and turn it into a jpeg. If it was a nice camera I would pay around $50 or so.
For an NXT camera sensor I would like it to take a picture when the program states. Then store it as a file on the bot. Lets you take it off when you plug it into your computer and will store it as a jpeg. i would pay around $50 or so looking the cost of the other camera sensors out there. Hopefully the design of the product would be as nice as the starting sensors that the NXT came with.
Post a Comment