can simulink use OCR to detect a 'Letter T' from a live video stream obtained from a camera on a UAV using PixHawk ? Or OPENCV is more preferred ?

Hi, as part of my final year project I am assigned to design an Autonomous Vehicle that has to fly at specific way points and drop payload on a red box with a letter T inside it (for more info see Page 27 of http://www.imeche.org/docs/default-source/agenda-documents/uas-challenge---competition-rules---issue-5-1.pdf?sfvrsn=2)

My team has decided on a hexcoptor using PixHawk as avionics. I already have downloaded the PixHawk support for SIMULINK. But I am confused as for the image recognition bit. I know that SIMULINK has live video streaming in computer vision toolbox but can I use OCR on this letter T or Do I have to use OPENCV instead.

Answer 1

[Dima Lisin]

Doing OCR for this seems like an overkill. Try detecting the marker using combination of color and local features, like SURF or MSER. See the Color Threshold App for color-based segmentation, and the Object Detection in a Cluttered Scene Using Point Feature Matching example for local feature matching.

[Mahed Javed]

Hi Dima,

Thanks for the answer. I've had a look at SURF options. I know it detects the specials features from the object to be located. So this would mean i would have to make a replica of the marker beforehand. That's alright but I just have got two questions :-

a) How can I detect from MATLAB the marker after I run my SURF code, I mean my input is in the form of a video cam which is onboard the drone. The example shows how you can do from an image. How can I achieve the same from a live video feed ?

b) Also I am still quite unsure how to put in ColorThresholdApp in it ? Meaning I dont know how can i use it here ? Or is it even going to help in the programming part ?

Thanks