Fast Logo detection on a moving target and control gpios

Project Discussion
1

Hello,
Iam new to the openmv system. I have experience with raspberry pi and opencv. And embedded systems

Iam doing an industrial project on Logo detection of a rectangular metal piece on a conveyor belt.
The metal piece moves at 5cm/s. I tried the keypoint detection / edge detection + blob detection. Using Ovi camera sensor
These algorithms work at small speeds of movement.but not on my required speed
The fps is automatically adjusting and the image capture is blurry( i capture in grayscale QVGA).
After detection i want to control a pneumatic actuator using gpio( thatz not a problem)

Can anyone suggest a good method for accurage detection of logo with faster processing speeds. Since iam working only on RAM. Saving images to device slows down the system. Also i get a feeling that my current algorithms in openmv is not running at full potential.
.
Any information regarding this topic would be extremely helpful.

2

Hi, if the image is blurry you should adjust the camera focus.

As for the method. Why not just blob detection by itself? Is the logo colored?

1 Like
3

By logo i mean a punched impression inthe metal. The focus is adjusted. I think the blurring is actually motion blurr.

4

The blob detection is not accurate…iam actually working on it.i found out the area of the metal peace and filtered out. The moving metal peace is detected. The impression ( logo) inside is actually detected as some contours. Iam actually adjusting the roi and other parameters and testing the device😇

5

Yeah, so our image library needs a lot of work.

Luckily, that’s going to happen in the future. I am working on a plan to go full time on OpenMV so I’ll be able to deal with a lot of things lacking in the vision library to bring it up to parity with OpenCV for many computer vision tasks.

1 Like
6

Where can i find the backend library files other than examples in windows pc. It may help with Stm32/ hal / fpu firmware level codes. So that we can fine tune the system

7

See our GitHub. The entire firmware and image processing library is open source. You can build the firmware and reflash the device.

1 Like
8

Ιf the image is still blury after adjusting the focus you may have too much exposure time for the image. Please keep it under 1ms.
You can try tamplete match if the label isn’t rotating too much too.

1 Like