Author(s)

Song Zhengcheng

Corresponding Author

Song Zhengcheng

Abstract

For the problems existing in the traditional design of feature extraction, image recognition and classification, this paper proposes to use CNN (Convolutional Neural Nets) for preliminary image recognition based on single feature of color or texture, and establishes a basic probability assignment according to the output of CNN as evidence. Then the recognition results of CNN are fused by DS evidence theory, and finally the image recognition results are obtained. The simulation results show that the average recognition accuracy rate of DS-CNN can reach 92.19%, which is much higher than that of single feature recognition and simple combination recognition, and it improves the stability and accuracy of image recognition with a more reliable result.

Keywords

Color Feature, Texture Feature, Neural Nets, Evidence Theory