Paper on Visual Players Detection and Tracking in Soccer Matches

This manuscript presents object detection and tracking in video clips of soccer matches. Background subtraction based framework on pixel energy is proposed to detect objects in the input video clip with varying light conditions, high frame rates, and real-time processing. Unsupervised clustering is then used classify detected objects into various classes. A stochastic approach based on maximum a posteriori probability (MAP) is proposed for object tracking. 1. Comment on Motivation and Novelty The proposed scheme is well motivated as player tracking in games such as soccer match video with high frame rate and varying light conditions in real-time is challenging task and the proposed scheme is intend to address this problem with very few constraints. Submitted manuscript presents a marginally novel framework to achieve promised goals. 2. Comment on Methodology Methodology proposed in the submitted manuscript for multiple object tracking seems incorrect, this might be due to typos. For example, consider Eqs. 1 and 2, these equations seem incorrect in their current form, that is, μ̄ = αμ + 1 − α + μ̄t−1 (1) σ̄ = ασ + 1 − α + σ̄t−1 (2) I guess factor 1−α is simply adding bias to the estimated mean and variance values which is desirable in parameter estimation. One possible solution to fix this problem is by using following modified versions, μ̄ = αμ + (1 − α)μ̄t−1 (3) σ̄ = ασ + (1 − α)σ̄t−1 (4) Following are the desirable modifications in the proposed framework: • Similarly, Eq. 4 also need to be modified, as we are dealing with discrete values, therefore it is reasonable to use summation instead of integration, that is, E (x, y) = ∑ t∈ W | I(x, y) − BC(x, y) |2 (5) • It seems that there is typo in Eq. 5 as well, I guess it should be modified as, BF = { [μ (x, y)] if E (x, y) > th(W ), [0, 0] otherwise (6) Here are few concerns about Eq. 5 (in the submitted manuscript), – What is φ? I guess a vector of zeros. – What is th(W ) and how it is determined? • Authors also need to modify Eq. 6 to cover all logical possibilities over E (x, y), th(W ), and BF (x, y), as in its present form it does not cover all logical possibilities over E (x, y), th(W ), and BF (x, y). • What is the significance of Eq. 7, authors need to add some description for Eq. 7 to enhance readability of the submitted manuscript. • Representation of Eqs. 10 and 12, can be improved by using vector notation. • Finally, in Eq. 11 how disturbance matrix, Σ, is estimated? IEEE Fifth International Conference on Advanced Video and Signal Based Surveillance 978-0-7695-3341-4/08 $25.00 © 2008 IEEE DOI 10.1109/AVSS.2008.79 337 Authorized licensed use limited to: UNIVERSITY OF MICHIGAN DEARBORN. Downloaded on August 02,2010 at 19:09:18 UTC from IEEE Xplore. Restrictions apply. 3. Comment on Experiments Experimental results presented in the submitted manuscript show that the proposed scheme can successfully track multiple-objects (players or group of objects) from diverse data set. It would be better if authors include some comparative performance analysis with existing player tracking (group of objects) schemes proposed in [2, 3, 1] (if possible). 4. Comment on Presentation Presentation of the submitted manuscript can be improved by careful proof reading and using distinct variable representation throughout the manuscript.