Research On Labeling Household Tool Parts In Affordance Area

The affordance of the tool is a tool provided by the behavior possibility,it is the value of the tool.The affordance detection and marking belongs to the field of object recognition,aiming at apperceiving the right area of the tool parts.The existing detection algorithms are effected by background and the light intensity,so the affordance marking is not accurate enough,this paper mainly focuses on tools of affordance detection and labeling algorithm,the research main contents are as follows:First of all,existing edge detection algorithm can not extract the affordance edge,so put forward a kind of affordance edge detection algorithm based on structured random forest,the algorithm for modeling the affordance of edge model in the offline stage;with the help of rough-to-accurate thought,learn different affordance edge threshold,and obtain the objective affordance rectangular area.The experiment results show that the algorithm can extract the edge with high precision for different tools.Secondly,objective affordance rectangular area exists noise in background,so this paper presents a algorithm based on the objective calculation of region labeling algorithm,the algorithm uses the related knowledge of the probability function of Markov absorption to calculate saliency in objective affordance rectangular area,and then uses the threshold filters the background noise,and marks the affordance area.The experiment results show that the algorithm can work successfully in a single scene and can label the correct location of the affordance region.Finally,in view of the significance and robustness of eage feature in image segmentation,put forward a kind of affordance region labeling algorithm based on edge linking.First,using the affordance edge model to detect the affordance edge,determining the affordance of the retangular area.Then,the algorithm finishes edge connection task in the objective affordance rectangular area,forms closed areas.At last,fill the closed areas,complete the affordance of region labeling work.The experiment results show that the algorithm can get good running efficiency and accuracy when the affordance parts are used in a single scene.