基于组件树结点属性的静态手势分类

Hand Posture Classification Based on Component Tree Attribute

提出一种基于图像组件树结点属性特征的静态手势图像分类方法。对手势图像建立组件树,并对原始组件树做简化处理以保留重要的组件,提取组件树结点的属性特征,建立数据集,使用全连接神经网络训练分类器,最后对该分类方法进行实验验证。

Proposes a hand posture image classification method based on the attributes of image component tree nodes. Component trees are built for hand posture images and simplified to retain important components. After that, the attribute features of the component tree nodes are extracted. A data-set is established from the extracted features to train a fully-connected neural network classifier, verifies the classification method with the trained classifier.