Let's Make Robots!

Pinhole camera - technique used for navigation?

Playing around with a small project, of a time-lapse-recording that should track a cloud in the sky, I got to think of Pinhole cameras - and I was thinking if any one had tried to play with the technique

- and instead of photographic negative on the backside, have a few light dependent resistors of some form?

A maze solving robot might see the walls ahead that way. And if you where tracking on IR, you might be better off with a line of sensors behind a pinhole, than scanning from side to side?

Other ideas?

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

I did some experimentation with a very simple pinhole eye recently. The basic idea was to view through a very small hole, resolve on a CdS cell, mount the sensor on a pan and tilt servo setup, and read a grayscale world one pixel at a time. This approach worked better than you might imagine. I printed out test patterns on an inkjet printer and was ultimately able to make recognizable low resolution "scans" at a distance of about 6 inches. I learned a couple of things along the way. Pin hole cameras try to make the hole in material which is as thin as possible in order to get a wide field of view. For my purposes I needed the opposite and worked to get very small holes in very thick materials in order to achieve a narrow field of view. Otherwise you would need hundreds or thousands of sensors to resolve anything besides light and dark. The other thing I quickly discovered is that CdS cells are woefully slow for this purpose, taking on the order of 25-100ms to respond to changes in light. I began to work with various phototransistors but had trouble finding an ideal part. I've since become distracted with some other projects but I do intend to pick this one back up sometime. In my research I found that early experiments in television used some novel techniques with pinholes and rotating mechanisms to capture images using a single sensor. For inspiration you might try searching for analog or mechanical television and also Nipkow or Nipkov disk.

Nipkov's system used only one sensor, but nowdays sensors are much cheaper so could put several of them in line and spin the line just like one of those POV (aka propeller) clocks... this way you can get a crude 2d image. The interesting thing about this image is that you will get very good resolution in the center and a lower resolution for the perifery, similar to a human eye. The best thing about this kind of camera is that you can use sensors that hook up directly to the microcontroller and you can get a low resolution image directly into the RAM. You are definately on to something Frits...