集成成像系统实现三维物体旋转不变分类识别

提出一种基于集成成像系统和综合判别函数(SDF)实现三维物体旋转不变分类识别的新方法。方法利用集成成像系统获取旋转三维物体的单元像阵列图像,应用图像中各个单元像记录的三维物体信息之间的高关联特性,结合SDF,实现了三维物体的旋转不变分类识别。与已有方法相比,本文方法减小了识别过程的数据量,降低运算复杂度,提高了识别效率,且不受物体旋转角度的限制。针对多类旋转三维物体,利用本文方法实现了旋转不变分类识别,验证了方法的有效性。

Three-dimensional object rotation-tolerant classification and recognition base d on integral imaging system

A novel approach of three-dimensional (3-D) object rotation-toleran t classification and recognition using synthetic discriminant function (SDF) is proposed based on integral imaging (II) system.In this approa ch,elemental image arrays of 3-D rotational object can be obtained by the II system,and 3-D rotation-tolerant classification and reco gni tion is realized by using the high relevancy characteristic in the information of 3-D object collected by the elemental image arrays and SDF.T he intention of SDF is to find a filter function that can obtain the same correlation output if rotational objects are input.But gene ral SDF needs more data to improve the recognition accuracy, which result in greater computational complexity and lower practicability.Meanw hile,high relevancy of perspective elemental image array of II system can overcome the above shortcoming of SDF.In compari son wit h other approaches,by introducing SDF into the field of 3-D object recognition,and combining the merits of II and SDF,the approach in this paper solves the contradiction between recognition accuracy and data amount,provides a novel recognition theory and therefore can greatly reduce the data amount during the recognition process and enhance the recognition efficiency.In addition,this approach can be applied for random rotational angles.Rotation-tolerant classification and recognition has been re alized for several kinds of rotational 3-D objects in this paper and the effectiveness has also been validated by experiment.