A Real-Time Video Analysis With an Omni-Directional Camera for Multi Object Detection Using The Hough Transform Method

Many computer vision applications require multi-object detection, such as robotics, surveillance, and autonomous vehicles. Hough Transform (HT) is a popular method for object detection in digital images, but its application to real-time video streams is limited. In this study, we propose a method for multi-object detection in the form of ball and goalpost using the HT algorithm on real-time video captured by an Omni-Directional camera. The proposed method consists of two main steps: the first step involves applying the HT algorithm to the video stream to detect potential object locations. The second step involves implementing a clustering algorithm to group candidate locations into object instances. Our method includes a pre-processing step to filter out the color of specific HSV objects and remove noise. Experimental results on real-time video can detect objects in the form of balls with an accuracy rate of 72.19% and goalpost with an accuracy rate of 80% in some random video streams with an Omni-Directional camera. So that in the future Omni-Directional cameras can become a valuable tool for various computer vision applications.