How to run Hough transform in real time?

Hi everybody

I would like to detect single object (circular) from the video in real time. I would like to do this by using Matlab. I am aware that Hough transform is a solution for this.

My question: Can I do this by using Hough transform in real time ( up to like 10 to 15 fps)? Is there any other method that you can suggest me?

Thanks!

Answer 1

[Prabhan Purwar]

Hi,

You may make use of GPU accelerated Hough transform algorithm to detect a circular object in the acceptance rate of 7 fps (real-time).

GPU Accelerated Hough Transform involves the following process:

The image taken from the webcam is preprocessed with a Median filter followed by Sobel edge detection and a thinning algorithm. Then, a Generalized Hough Transform is performed using a custom, user-defined shape.

Object Detection can be classified into three broad categories:

(1) Image processing/Computer Vision - Color Thresholding, Blob Analysis, Histogram of Gradients, Speeded-Up Robust Features.

(2) Machine Learning- Cascade Object Detector (Viola-Jones Algorithm), Aggregate Channel Features (ACF)

(3) Deep Learning - YOLO v2, R-CNN, Fast R-CNN and Faster R-CNN

Refer to the following links for further information:

• https://www.mathworks.com/matlabcentral/fileexchange/71522-designing-object-detectors-for-real-case(Design Object Detector for Real Cases)
• https://www.mathworks.com/matlabcentral/fileexchange/40022-how-to-detect-and-track-red-colored-object-in-live-video(Detect and Track Red Colored Object in LIVE Video)
• https://www.mathworks.com/matlabcentral/fileexchange/52183-simple-demo-of-hough-transform-implementation(Simple Demo for Hough Transform Implementation)
• https://www.mathworks.com/help/images/ref/hough.html(Hough Transform)
• https://www.mathworks.com/matlabcentral/fileexchange/25157-image-segmentation-tutorial(Image Segmentation Tutorial)

[engineer]

Thank you very much. I will look into the links you provided.