三维物体识别的心理表征:角度依赖还是角度独立

The Mental Representation on Three-Dimensional Object Recognition:Dependent or Independent on the View

"物体识别是角度独立还是角度依赖"的问题一直存在着争议。以物体为中心理论认为是角度独立,以观察者为中心理论认为是角度依赖。研究参照物体识别的"小几何体"思想自制实验材料,采用启动范式,在两个实验中通过考察结构信息和视图信息对物体识别的影响,揭示物体识别的心理表征。结果发现:(1)形状信息和类别空间关系信息影响物体识别,表明以物体为中心理论对物体识别表征机制解释的合理性;(2)角度信息可能独立于形状信息表征,深度旋转方式下的识别成绩优于平面旋转,表明以观察者为中心理论能够更好地解释三维物体识别的深度旋转现象。研究结果为物体识别理论的融合提供了实证依据。

Object recognition refers to a decision about whether an object has been seen before.Recognition involves generalization across shape,size,spatial relation,location,view and illumination.The view has been the focus of the debate in many recent object recognition studies.The object-centered theory assumed that object recognition is a qualitative representation of an object in terms of its three dimensional primitives（for example,geometrical ion） and their spatial relationships,independent on the view.By contrast,the view-centered theory posited that object recognition is dependent on the view.It is view-specific,or represented in an egocentric frame of reference.It might contain information about illumination,colour,material and depends systematically on experience with specific views of an object.We decided to investigate the roles of structural information and view information in object recognition,with specific emphasis on the potential separability of the shape and the rotational ways of the view in object perception,as predicted by the view-centered theory.In the current study,made 3D objects as experimental materials according to the idea of Biederman＇s small geons;used the priming paradigm;adopted identify task;participants were required to judge whether two objects are the same or not.The effect of the 3D object recognition on the PC（part-changed） or RC（relation-changed） distractors is studied in Experiment 1.A 6（view：-30°,0°,30°,60°,90°,180°） × 4（compound mode：DoDv,DoSv,SoDv,SoSv） × 2（task：indirect,direct） mixed design is utilized in Experiment 2.The rotations were clockwise by 90°and 180°in picture-plane level,and the rotations were counter clockwise about the vertical axis by-30°,30°and 60° in depth.In the indirect task,those participants were told to respond ＇same＇ if two pictures appeared at any view.In the direct task,those participants were told to respond ＇the same＇ if two pictures appeared only at the identical view.The results confirmed our hypothesis.In Experiment 1,the speed of object recognition in repeat prime was faster than those in the part prime and the spatial relation prime,there was a significant difference between the RTs of the part and spatial relation prime.In Experiment 2,ANOVA revealed the main effect of the RTs of the rotational view,compound mode and task type,as well as a significant view-by-mode,mode-by-task interaction.Especially,the RTs in the SoSv mode are shortest.That is,the view prime effect was found in two tasks.The size of priming effect is affected by the rotational ways of the object.Based on these results,we concluded that（1） the speed of 3D object recognition was influenced by the shape of part and the categorical spatial relation,which suggested that the object-centered theory was rationality.（2） The view information was independent on the shape information,and the depth rotation was better than the picture-plane rotation,which suggested that the depth rotation phenomenon can be better explained by the view-centered theory.The findings not only provided experimental evidence for object recognition,but also shed light on the object theory.