论文部分内容阅读
为满足人机共融环境下机器智能对工具功用性认知的需要,模拟人类自底向上的认知方式,设计了一种基于部件功用性语义组合的聚类方法,来对家庭日常工具进行表示与建模.首先,设计了工具功用性部件边缘表示方法并基于结构随机森林加以建模.然后基于功用性部件组合思想,设计了高层语义空间上联合各部件显著度的工具整体表示方法并采用聚类方式构建工具功用性字典.在线检测阶段,联合测试样本各功用性部件的显著度,利用其与工具功用性字典的距离残差对工具进行分类判别.在实验中,将7种功用性部件组合聚类形成5类工具,当各类工具选取不同核值时,分类精度可达90%以上,即使各类工具的核值固定为3,分类精度也在85%以上.实验结果表明,相较于传统的特征表示方式,功用性语义的加入使机器人深化了对工具功能的认知,基于功用性部件组合的字典表示使得家庭常见工具的分类精度和效率明显提升,且实现了工具间功能相似性测算和最优替代工具查找.
In order to meet the need of machine intelligence to recognize the utility of the machine in the environment of human-machine communion and to simulate the bottom-up cognitive mode of human beings, a clustering method based on the functional semantic combination of components is designed to carry out the daily household tools Representation and modeling.Firstly, the edge representation of tool utilizable components is designed and modeled based on structural random forest.Then, based on the idea of functional component combination, the overall representation of the tool combining the saliency of each component in high-level semantic space is designed The clustering method was used to construct the utility dictionary of tools.At the stage of online testing, the saliency of each functional part of the test samples was united, and the tool’s classification was discriminated by using the distance residuals between them and the utility dictionary of the tools.In the experiment, When the different types of tools choose different nuclear values, the classification accuracy can reach more than 90%, and the classification accuracy is over 85% even though the core values of all kinds of tools are fixed at 3. The experimental results show that , Compared with the traditional feature representation, the addition of functional semantics makes the robot deepen the understanding of the function of the tool, and the dictionary based on the combination of functional components Was a common household tools, precision and efficiency improved significantly, and realized the similarities between functional tool measurement and tool to find the best alternative.