Research on target localization and recognition based on binocular vision and deep learning with FPGA

Abstract

Agricultural picking robots have ushered in a good development period, and the research on the related technologies is becoming more and more practical. Based on this, this paper designs a crop identification and localization method based on binocular vision and deep learning supplemented by FPGA hardware platform. This paper firstly introduces the necessity of the technology for recognition and localization, then introduces the theoretical knowledge of camera calibration, stereo matching, and convolutional neural networks, etc. Finally, the implementation of relevant algorithms and the system architecture model are carried out. The method designed in this paper can provide technical support for agricultural picking robots, which can contribute to the development of picking robots.