基于有向图的足球视频中球员分类算法

Player classification algorithm based on digraph in soccer video

基于传统的直方图球员分类方法由于缺乏描述图像颜色的空间信息而造成分类误差,而且该方法需要先验的模板信息。为此,提出一种基于有向图的足球球员的分类方法。首先,利用HSV模型中主颜色方法提取候选球员,并利用等面积矩形划分策略对图像进行分块;其次,对子块的HSV颜色空间进行量化,将统计直方图作为颜色特征,然后通过颜色特征计算图像之间的距离,并利用距离矩阵生成对应有向图;最后,通过对有向图的顶点分类实现球员的分类。实验结果表明,提出的方法在没有先验模板信息的条件下,能够有效地解决处在分类边界上的球员分类问题,正确率达到98.23%;与传统方法相比,具有更好的分类效果。

Player classification method based on the tradition histograms lacked color spatial information,which led low classification performance, in addition, it needed priori template intbrmation. According to this, the paper proposed a novel player classification algorithm based on digraph. Firstly, the proposed method extracted candidate through the main color in HSV model and adopted equal-area rectangle partitioning strategy to partition the image. Secondly, it quantized the HSV color space in each block and extracted color histogram as color features. Thirdly, it calculated the distance among images by using color features and then generated a digraph based on the distance matrix. Finally, it implemented the player classification by classifying the vertexes of digraph. Experimental result shows that the proposed method can classify the players positioned around the classification boundary in an effective way with an average accuracy of 98.2%. Compared with traditional method, the proposed method has a remarkable promotion on classification effect.