A Recognition Method for Multi-object Information Based on Multi-source Data Fusion

Abstract

The accuracy of object recognition is difficult to be improved by single information acquisition, we propose a recognition method for multi-object information based on multi-source data Fusion. By analyzing the high-level semantic features of RGB images and Depth images, feature fusion module is added, then, the calculation of parameters of the model is reduced based on the idea of residual learning. Combined with the GRU recursive neural network, a tighter feature sequence is to generated, which improved the accuracy of RGB-D object recognition. Finally, improved method has been experimented on multiple public data sets, the results show that the object recognition method in this paper integrates depth information, Compared with single RGB image, the recognition accuracy is significantly improved; Compared with other RGB-D-oriented deep learning methods, the recognition accuracy of the method in the article has been significantly improved by at least 2.5% in 2D3D dataset.