I'm working on a picture project and I need your help The project is as follows: I want to estimate the position of people in the real environment, I have used a camera to do this, I have detected people in the image with the help of the Yolo algorithm (python / opencv ) and I have determined the coordinates of their bounding boxes, now I want to determine those coordinates in the real environment. Do I have to use a matrix camera calibration? What do you suggest to do this? thanks
Asked
Active
Viewed 276 times
0
-
Yes, you need to find the bottom of the people in the image and assume that is ground level. Then you need to use the camera model to project rays from the focal point, through the image at the foot of a person and intersect that with the ground (Z=0 plane) to find the corresponding real world X and Y coordinates. Other approach is discussed at https://stackoverflow.com/questions/61674231/how-to-calculate-distance-between-two-person-using-python-opencv/61675660#61675660 – fmw42 May 08 '20 at 20:22
-
1See also https://zbigatron.com/mapping-camera-coordinates-to-a-2d-floor-plan/ and https://www.fdxlabs.com/calculate-x-y-z-real-world-coordinates-from-a-single-camera-using-opencv/ – fmw42 May 08 '20 at 20:29