Synchronizing footage from two cameras with an OpenCV algorithm 

Hi, Cemhan. Thank you for your interest and feedback! We intended to use this algorithm as part of a bigger computer vision system for object tracking, for example, to track people in a store. Since one camera may occasionally lose the tracking object, the second camera would be able to continue tracking it. Another use case is to have multiple cameras distributed across a building with each spot being covered by at least two cameras. We did a series of researches in order to understand how accurately the algorithm can track different objects. We used the OpenCV library for fish eye calibration, stereo calibration, and stereo vision. The algorithm itself is fast and can run even on a Raspberry Pi device. However, if you want to combine it with object detection models, you may want to use more powerful hardware, such as NVIDIA Jetson. Of course, you can compute ‘global’ coordinates, but for our research it wasn’t necessary: we only needed to map objects from one camera to another. The blackbox you mentioned was originally used as a ‘sandbox’ where we could flexibly adjust the position of all 8 points to understand how it affects accuracy. It consists of wires extracted from a twisted pair cable and does not contain any electronic parts. If you are interested in experimenting with this algorithm, take a look at the official OpenCV example: docs.opencv.org/4.x/da/de9/tutorial_py_epipolar_geometry.html