วิธีการจัดหมวดหมู่ผู้เล่นคนเดียวกับผู้เล่นหลายคนภายในสนามฟุตบอล

Abstract

Player position analysis from offside position which is an important event that affects outcome of football match. An important step to player position analysis of each player is a soccer team classification. The classification of football players on the field is a very challenging problem. เท this research, we propose a method for classification single player and multiple players on soccer field using image processing techniques on broadcast video obtained from the internet. เท order to classify single player and multiple players on the soccer field. There are 5 steps in total, consisting of 1) image preprocessing, 2) player’s area segmentation 3) edge detection, 4) find features of image 5) classification. Firstly, segment foreground objects, the soccer ground field is removed by comparing the green color level with a threshold. The other part is the player detection process. The result is an image with players detected on the football field. And get the player's properties out with area, coordinates and number of pixels. Then import into the process of player’s area segmentation and bringing the properties after that remove those positions from the image to get the result of the process. Next edge from the previous process into the process find features of image to store the dataset into the classification process. From the experimental classification process used Decision Tree classification, in this process was performed on a sample of about 400 images using 6 features that is the y-axis variance, the x-axis variance,ง the ratio of x-axis variance to y-axis variance, pixel length, pixel width, and the ratio of pixel width to pixel length. The data set was used to verify the data by K-Folds cross validation, which divides the data into 5 folds equally. There were 80% of training data and 20% of testing data, which resulted in accuracy is 85.30%, sensitivity is 85.53% and specificity is 85.09%