2
$\begingroup$

For a University project I have to use computer vision to detect small drones within 40 feet. I know there exists a pixycam for this purpose, but I was not happy with it, when I used it for CV.

I have a normal digital camera which is 16 Megapixels (pic & video), which I don't use anymore. Before I dissect the camera, I was wondering if it is practically possible to train this digital camera for computer vision - detecting small flying drones. Any thoughts on this - using a digital camera for CV?

Thanks

$\endgroup$
  • $\begingroup$ Can you please specify the requirements with details ? Which 16MP camera is it? How many drones do you need to detect? How small are they? What aren't you 'happy' about? What is the frame-rate? (can be used for sparse-detection) Sample images of 16MP camera and PixyCam? I am eager to help you. But, I am sorry. I can't unless you post details. $\endgroup$ – Prasad Raghavendra Mar 29 '17 at 6:38
  • $\begingroup$ @PrasadRaghavendra . It is a Cannon A2300, 16 megapixel digital camera. I need to detect a 1 or 2 drones max. The pixy Cam is only 1 megapixel and doesn't work very well in low light & finds it difficult to differentiate between dark colours and doesn't detect the drone when there is something of similar colour in the background. Im not sure of the frame rate. But I was wondering if we can program or write algorithms that can detect make the Canon detect the drone like how it detects peoples faces & smile $\endgroup$ – Stanford Mar 30 '17 at 20:04
1
$\begingroup$

If the university project is not directly about making a drone detection system using visual technic, it may not be worth the effort to connect the "camera" to the computer. (You may have to install new drivers, etc).

However, technically it's quite possible to run a feature-detection-code once you can integrate the camera to that code. There are several open source and free (as in freedom) image / video processing algorithms that are able to detect features. See GitHub for some examples.

However, they are not tuned for drones. You could help out others by sharing your experience; be it in academic studies or open sharing.

The problem with visual detection is that it will not detect drones at poor visibility conditions, night, fog etc. Implementing an IR camera could help a little.

$\endgroup$
  • $\begingroup$ Yes, we do have acoustic detection too. But we wanted CV to work well too, in low light conditions. The digital cam is better in lowlight compared to the pixycam $\endgroup$ – Stanford Apr 1 '17 at 14:56
1
$\begingroup$

First make sure you have the camera's drivers available for the operating system you'll be running your code on. If drivers are available for your platform check what resolutions and fps(s) you can read from the camera using the driver.

Once you have those things in place you'll be able to do anything in cv that will accept raw (YUV or YUV2 or whatever your camera output's via the driver).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.