Research On Detection Algorithm Of Affordance Of Household Daily Tools For Home Service Robots

Under the background of the advancement of the research on artificial intelligence and the aging of the population,the requirements for the intelligentization of service robots are getting higher.How to serve humans more intelligently becomes an important part of the research field about home service robots,among which,for household daily tools cognition requires robots to accurately identify and use tools.According to the requirements of tool affordance cognition accuracy and real-time,this paper studies from the perspective of tool affordance.Based on machine learning algorithm,robots can understand the affordance of various tools through independent learning.The main research as follows:First of all,the household daily tool components have a complex structure,and the feature representation method based on shape and color is difficult to fully represent the features.For this problem,a Local Binary Pattern based on Low Rank Matrix Recovery is proposed.Firstly,the Local Binary Pattern features of the Low Rank Matrix Recovery household daily tool data are extracted,and then the sparse representation model is constructed.Finally,the dictionary learning algorithm is used to solve the affordance model and realize the affordance detection and classification of the household daily tool components.Secondly,on the basis of the functional characteristics learning,this paper proposes a tool affordance detection algorithm based on Deep Belief Network and Support Vector Machine.The traditional Support Vector Machine classifier cannot be directly applied variety of affordance detection,a Multi-classification Support Vector machine is used for classification.Firstly,the particle swarm optimization algorithm is used to optimize the Deep Belief Network learning rate.Then,the improved Deep Belief Network is used to learn the feature data of the household daily tool.Finally,the learned features are input into the Multi-classification Support Vector machine for affordance detection.Finally,the household daily tool is composed of several functional components,and each component has different geometric features.The traditional feature representation and extraction algorithm is difficult to effectively characterize the affordance of the tool component.For this problem,a Deep Belief Network tool affordance detection algorithm.Firstly,the feature data of the household daily tool data is learned by using the Deep Belief Network,and then the Softmax classifier is added at the end of the network for identification.The results show that this method can achieve higher detection accuracy and at the same time has a certain improvement in runtime.