|
Acadêmico(a): Paola Adriano |
Título: ANALYTICS-BOL: UM PROTÓTIPO PARA GERAÇÃO DE ESTATÍSTICA EM UM JOGO DE VOLEIBOL |
|
Abstract: |
This work presents a prototype for generation of score statistics in volleyball game. The images were captured manually from videos made in a training session of the women's team of Blu Volleyball at the FURB gymnasium. Images were selected that contained the ball on the court, showing a point, then being subjected to several stages of image processing and, finally, generating statistics of how many times it fell in a certain quadrant of the court. The first step of image processing was to determine what the court was. First, the image was converted from RGB to the HSV color system. The V channel was used to delimit the court, being made the segmentation by thresholding and the noise removal by the morphological operators of dilation and erosion. From this, in the next step, the players are removed from the court so there is no possibility of them being confused with the ball. The third step was to discover the center of the court, to process the parts in isolation, limiting the area of search for the ball. In this process the Hough transform was used. From the left side of the block, the image was converted again from RGB to HSV, where the channel H was used to predict the location of the ball inside the court. After identifying the location, the percentage of how many times the ball appeared in a given quadrant was calculated. In total, 45 images were analyzed, where in 32 the ball was positioned on the left side of the court and in 13, it was positioned on the right side of the court. In the 32 images on the left side, in only two the ball was not found and consecutively computing the statistics in the wrong way. Already in the 13 images on the right side, the prototype worked correctly until the step of cutting half a block, but could not identify the ball correctly in either case. |
|
|