Explicit semantic events detection and development of realistic applications for broadcasting baseball videos

Abstract

This paper presents a framework that explicitly detects events in broadcasting baseball videos and facilitates the development of many practical applications. Three phases of contributions are included in this work: reliable shot classification, explicit event detection, and elaborate applications. At the shot classification stage, color and geometric information are utilized to classify video shots into several canonical views. To explicitly detect semantic events, rule-based decision and model-based decision methods are developed. We emphasize that this system efficiently and exactly identifies what happened in baseball games rather than roughly finding some interesting parts. On the basis of explicit event detection, many accurate and practical applications such as automatic box score generation and game summarization could be built. The reported results show the effectiveness of the proposed framework and demonstrate some research opportunities about bridging the semantic gap for sports videos.

Multi-modal highlight generation for sports videos using an information-theoretic excitability measure

“…In pitching scenes, a strong valley is usually found in the left hand side of this distribution [36], as shown in Figure 12c, due to the presence of the pitcher. If such a valley is found, C V = 1; else 0. iv) Player location condition (C P ) : From the video frame, a binary edge image is calculated using the 'sobel' method and image dilation [37] is performed.…”

“…To detect the pitching scenes, the following operations are performed on each video frame: the field pixels are detected using the HSV color space condition as in [36] and a H × W binary image is formed. Here, W and H denote image width and height in pixels, respectively.…”

“…In pitching scenes, the lower half of the image usually contains more baseball field pixels [36]. Thus, if the lower half has more than twice the number of field pixels compared to the upper half, we set C L = 1, else 0, which becomes: …”

A multi-modal highlight extraction scheme for sports videos using an information-theoretic excitability measure

“…To detect the pitching scenes, the following operations are performed on each video frame: the field pixels are detected using the HSV color space condition as in [11] and a binary image is formed. Fig.…”

“…3 (b) shows an example binary image from a pitching scene. A video frame has to pass four condition tests to be eligible as a pitching scene: (i) area ratio [11] lies in the range of 25% − 45%, (ii) lower half of the image contains more field pixels [11], (iii) strong valley found in the vertical profile of field pixels [11] (as in Fig. 3 (c)) and (iv) higher intensity occurs in specific block regions (7, 11, 10, 14 as shown in Fig.3 (d)) due to the presence of the pitcher and batter [6].…”