Longshot Football Detection and Tracking Using Mean Shift Algorithm

Kumar, Chitturi Vinod (2015) Longshot Football Detection and Tracking Using Mean Shift Algorithm. MTech thesis.



The soccer video analysis has a lot of importance these days commercially. There are many challenges in making the soccer video automatically analyzed. The automatic analysis is highly useful in analyzing player performance, computer assisted referee and tactics inference. In this thesis, we have proposed distinct methods to capture the ball movements there by ending which team has the control on the ball for most of the time. Distinct methods mean different methods for detection, tracking and to find the control. In the detection to identify the location of the ball, we have proposed a 9 step method. We have to execute these steps one by one. These steps include played detection, edge detection, moving object detection, conjunction, morphology, shape strainer, size sieve, dominant color extraction, neighborhood detection. These steps are mainly based on attributes of the ball. These attributes include color, shape, and size. The moving object detection step work based on the optical ow of the ball. In the tracking phase, we will select a region of interest in which we will nd the most probable location of the ball by maximizing the likelihood of multiplication of color histogram of location and ball. After finding the most probable location, we localize the ball in the region by mean shift tracking method. The mean shift tracking algorithm is the one which works base on weights derived from Bhattacharyya coecient to delineate the ball or target. Bhattacharyya coecient is naturally used to nd the similarity between two statistical samples by considering their probability distributions. In the control phase, we divide the frame into di erent clusters using k-means algo-rithm. In k-means, we start the iterative process by selecting random starting points.The image pixels which has color values close to the di erent centers will be clustered into different regions. We will nd the centroids or centers of these di erent regions and run the process again. We have applied this algorithm to our soccer frames got the results

Item Type:Thesis (MTech)
Uncontrolled Keywords:Longshot Football; Football Detection; Mean Shift Algorithm; CAM Shift Algorithm; K-Means Clustering; Control on Ball
Subjects:Engineering and Technology > Electronics and Communication Engineering > Image Processing
Divisions: Engineering and Technology > Department of Electronics and Communication Engineering
ID Code:7784
Deposited By:Mr. Sanat Kumar Behera
Deposited On:17 Sep 2016 14:48
Last Modified:17 Sep 2016 14:48
Supervisor(s):Meher, S

Repository Staff Only: item control page