DENSE MATCHING RELATION-BASED RGB-D OBJECT RECOGNITION METHOD USING ADAPTIVE SIMILARITY MEASUREMENT, AND DEVICE

摘要

A dense matching relation-based RGB-D object recognition method using adaptive similarity measurement, and a device. The method at least comprises: extracting convolutional neural network features of an object to be queried and a reference object (S102); on the basis of the convolutional neural network features in combination with RGB and depth information, performing dense matching on the reference object and the object to be queried (S104); on the basis of the result of the dense matching, measuring similarity between the reference object and the object to be queried (S106); and classifying the object to be queried on the basis of the similarity between the reference object and the object to be queried (S108). The method and device at least partially solves the technical problem of how to improve robustness of object identification.