Wednesday, 21 Feb 2018
You are here: Home News Mini Projects Progress

Subscribe to our Newsletter

How many eyes has a typical person? (ex: 1)

Twitter Feed

Mini Projects Progress PDF Print E-mail
Written by University of Reading   
Friday, 04 April 2014 17:00

Camera Calibration


In the latest development, new features were introduced to improve the robustness of the camera self-calibration algorithm. In order to improve the calibration accuracy, the user had to provide vertical lines from structure and identify any areas which have occlusions.

In addition, three algorithms were introduced to minimise the influence of noise on the calibration. The first one was a shape analysis over the detected blob. Such analysis excluded the blobs that do not represent a person with both legs visible together.

The second algorithm is implemented to eliminate the image local noise. The last method replaced the Least Squares Fitting algorithm with the RANSAC fitting algorithm for better performance. All these modifications improved on the previous results significantly as shown in the Figure.