基于可变形部件模型的台标识别方法
背景变化复杂、部分台标相似度高、拉伸变形等因素增加了台标识别的难度，降低了识别的准确率。为此，提出了一种鲁棒的基于可变形部件模型的台标识别方法。依据台标特性，利用合适的颜色特征对可变形部件模型的特征进行了改进和增强；利用隐式支持向量机和隐式线性判别分析技术加速台标识别模型训练。为了弥补可变形部件模型的不足，设计了一种基于加权部件的计算方法，提出一种新的可靠机制进行准确率评价。实验结果表明，与基于方向梯度直方图和支持向量机的识别方法相比，该方法具有更高的识别准确率，性能更加稳定。...
Saved in:
| Published in | 计算机应用研究 Vol. 34; no. 7; pp. 2202 - 2206 |
|---|---|
| Main Author | |
| Format | Journal Article |
| Language | Chinese |
| Published |
中国劳动关系学院 计算机应用教研室,北京,100048%青岛农业大学 理学与信息学院,山东 青岛,266109
2017
|
| Subjects | |
| Online Access | Get full text |
Cover
Loading…
| Summary: | 背景变化复杂、部分台标相似度高、拉伸变形等因素增加了台标识别的难度，降低了识别的准确率。为此，提出了一种鲁棒的基于可变形部件模型的台标识别方法。依据台标特性，利用合适的颜色特征对可变形部件模型的特征进行了改进和增强；利用隐式支持向量机和隐式线性判别分析技术加速台标识别模型训练。为了弥补可变形部件模型的不足，设计了一种基于加权部件的计算方法，提出一种新的可靠机制进行准确率评价。实验结果表明，与基于方向梯度直方图和支持向量机的识别方法相比，该方法具有更高的识别准确率，性能更加稳定。 |
|---|---|
| Bibliography: | Because of the complexity of the background, the high similarity of partial TV logo and the change of the shape of TV logo, it increases the difficulty of TV logo recognition and reduces the accuracy of recognition.Therefore, this paper proposed a robust TV logo recognition method based on the deformable part model （DPM）.First of all, based on the TV logo features, it used the appropriate color features to improve and enhance the features of the deformable part model.Secondly, it used the latent support vector machine （LSVM） and latent linear discriminant analysis （LLDA） technology to accelerate the train of the TV logo recognition model.Then, in order to make up the deficiency of the deformable parts model, it designed a calculation method based on the weighted parts.Finally, it proposed a new reliable mechanism to evaluate the accuracy of the TV logo recognition.Experimental results show that the proposed method has higher recognition accuracy and more stable performance compared with the recognition method |
| ISSN: | 1001-3695 |
| DOI: | 10.3969/j.issn.1001-3695.2017.07.061 |