Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

cheap camera telemetry

Status
Not open for further replies.

thaddeus74

New Member
Hi,
I'm looking for a reasonably cheap ($100-$150) camera that can send back it's telemetry data to a pc (a mac actually).

The idea is that I will buy several of them and set them at specific locations. I want them to send their images along with data such as yaw pitch and roll. Ideally focus depth and location as well, but I can do without that initially. I don't need wireless transmission, but I will need several hooked to the same computer at a time.

I don't need to be able to remotely control it's angle but that would be a nice to have as well.

Is there anything like that out there?

oh, and timestamps of the images would be good too, so I can correlate images from different cameras.
 

thaddeus74

New Member
This is for a simple skunkworks test on a scale model, so range will only need to be 10 feet or so, quality can be average...something like a 3M quality should be fine for stills. For video, i don't know...640x480 (vga quality) should suffice.
 

arhi

Member
note that camera that can send yaw/pitch will also have ability for you to control it

I do not see you getting one for 100-150$ (unless you get thousands at once).

timestamp from camera? camera will give you "current" view, so the timestamp need to be added on the receive side, not on the camera side.

there is bunch of specialised hw for attaching many of these cameras to a computer, some are PCI cards, some are USB boxes, I saw PCI cards that can connect to some 20 remote operating cameras (each one can change yaw/pitch/zoom). I know there are ones that can take more but never saw one. You use special software that talks to that "pci card / usb box". The "device" can monitor all feeds simultaneously and trigger interrupt when picture changes on one of the cameras. The software can view up to 9 (the version I saw) cameras in the same time (any of them) but can record stream from all of them at once hence, this limitation is probably only due to display size. The software can be set to put the "one with changes" in front windows etc... pretty nifty stuff but VERY EXPENSIVE..

now, each of these cameras (wire connection) was ~200E. There was ability to go wireless but the breakout box was ~100% more expensive and each camera was over 350E...

the simple ip camera (320x240 - 640x480 - 802.11b) ranges from 150 to 500$ depending on the quality of the pic, and there is no yaw/pitch control nor reading

btw, what is the point of reading yaw/pitch if you cannot change them ?? if camera does not have yaw/pitch control the only way to change it is by hand on the spot, in order for camera to know it's orientation and elevation it would have to have compas and gyro or accelerometer inside ?! what would be the point ?
 

dknguyen

Well-Known Member
Most Helpful Member
Do you mean something like this?

FlyCamOne2 Micro Video Camera for RC Aircraft from Hobby Lobby!

Pan-tilt is something you have to build yourself (for any camera in general, not just this one), let alone any yaw-pitch and roll data and time stamping. You would have to build the pan-tilt mount yourself and mount the camera on it. And for the orientation data you would also have to come up with something yourself. Why? Because a camera with any one of those things (except for time stamping) was made for UAVs, which probably means it was made for military...commercial if you're lucky. But either way, far beyond the cost of what a single person can afford.

Cameras that record their own yaw/pitch/roll are for localization and have internal IMUs. Considering the cost of an IMU is $2000+ for the cheapest ones (with most being $5000+) it's probably much more when built into a camera. Plus it's also heavy and a bit short-sighted to dedicate so many resources to such a menial task. I've never seen a camera with a built-in IMU to determine it's own orientation. It's expensive, redundant, heavy, and a waste of such an impressive piece of hardware. (An exception is stabilized cameras, but those just have gyros and aren't IMUs because they don't require absolute accuracy for long periods of time.)

It makes far more sense to use a single IMU to track the plane's orientation, and use that with the servo position feedback to figure out the camera's absolute orientation. THat's super easy to do...you need are some servos, a microcontroller, and an IMU (assuming you don't need to build the IMU). Your biggest problem is getting an IMU.

TIme-stamping is easy if the camera doesn't have it. If it's real-time video...no problem! If it's photographs, just make a circuit that records the time every time you click the shutter button. Then correlate the photos with the time afterwards since the camera should be storing them in order. But we're all lazy here and that takes some work.

Do you actually need to know when the photos were taken? Or do you just need to know that they were taken at the same time between cameras? THere's a real simple solution to that. Just use the same shutter signal for all cameras. Since the photos are stored in order it's very easy to know which photos were taken at the same time between the cameras.

But if you're building camera orientation+IMU feedback electronics, might as well add on time stamping since it's dead easy. A single MCU should be able to take the readings from an single IMU, and all the pan-tilt servo positions from multiple cameras and figure out all their orientations all at once every time the shutter is used. And since RC servos give no external position feedback, the only way to get it is to measure the position PWM signal. So the MCU might as well be the one sending those signal anyways. Then it could ensure the cameras are always pointing at a fixed angle relative to the ground (at least as much as the mechanics of the pan-tilt will allow).

There are so many things you could do.
1. THe cameras could be maintained at a fixed angle relative to the ground in which case you probably assume the cameras are always at X degrees relative to the ground in which case you don't need to read out the camera orientation feedback (if you are willing to assume the control loop is sound.) Only the control loops needs this info which saves on a lot of work with the interface.
2. Or you could take the more boring approaches like being able to change the camera's position in flight relative to the plane (not the ground) and the camera spits out it's own orientation data.
3. Or something like 1, except that you can adjust the angle that the cameras maintain relative to the ground in-flight. THen you'd need to somehow record what photo corresponded to which angle hold setting. But seems more like a gimmick to me unless you had a very good interface built. Because then it'd just become a pain in the ass to be setting the angle hold setting of multiple cameras in flight.

Cool stuff.
 
Last edited:

thaddeus74

New Member
The cameras are not going to be mounted on a model, they will be taking pictures OF a model. I don't necessarily need timestamps, I just need simultaneous pictures (as you suggested). I can of course fix the cameras and measure them by hand. That would be quick and simple...I was only looking for telemetry cameras to make additional shots easier to set up, without measuring every time I change a camera position.

The application I'm building is to take several camera angles and rebuild 3d models. I know that these already exist for stereometric cameras, but I'm working on algorithms that take a wider variety of camera inputs from initially unknown sources, and then uses machine learnig to piece together the 3d model.

Obviously, this is not really a robotics application, but I figured you guys have a lot of experience with small cameras, and I'm building a scale model for my initial testing, so I thought my question would be up your alley.

Assuming I get simple inexpensive webcams, are there any suggestions for taking somewhat accurate angle measurements?
 

dknguyen

Well-Known Member
Most Helpful Member
Oh, if it's not moving you can build a small with one MCU (that has ADCs built-in) and a 3-axis accelerometer IC (all of the ICs I have seen have analog output). THat will give you near perfect angular measurements. You just need to decide how you want to get the data back to the computer...I think RS-485 transceiver IC on the UART might be best but you will need to get an RS-485 to USB adapter to get it to work with your PC. RS-485 will let you have one cable communicate with multiple cameras rather than one camera per cable. You would need an RS485 transceiver for every MCU and a RS-485 to USB adapter to get it into your PC. You can also use RS-232 but you'd need one cable for every camera going into the PC and you'd still need RS-232 to USB adapters (either fully built adapters, PCB modules, or ICs as well as PC drivers, just like the RS485). All these essentially make the PC treat the USB as a serial COM port which is much easier to work with than native USB (for both the PC and the MCU).

The same MCU should also probably drive the servos in the pan-tilt unit so you can send commands for the camera to be oriented a certain way and the MCU will just do it for you by producing the PWM signals that RC servos use.

Oh, one more thing. You don't even need accelerometers if you know know where the pan-tilt unit is mounted in the first place. You should know the position of the RC servo since you are the one sending the position signal to it, even if you they don't have an external position feedback signal.
 
Last edited:

thaddeus74

New Member
actually it looks like there are a variety of webcams with tilt pan and zoom that should work fine and for under $200 a piece.

thanks for all the input.
 

micronetric

New Member
Look for a old cam corder and figure out its circuits to disassemble and use as you have to. No need in the exterior casing.
 
Status
Not open for further replies.

Latest threads

EE World Online Articles

Loading
Top