Research on computer vision enhancement in intelligent robot based on machine learning and deep learning

Abstract

The stable operation of intelligent robots requires the effective support of machine vision technology. In order to improve the effect of robot machine vision recognition, based on deep learning, this paper, under the guidance of machine learning ideas, proposes a target detection framework that combines target recognition and target tracking based on the efficiency advantages of the KCF visual tracking algorithm. Moreover, this paper designs a vision system based on a high-resolution color camera and TOF depth camera. In addition, by modeling the coordinate conversion relationship of the same object in the camera coordinate system of two cameras, the projection relationship of the depth map collected by the TOF camera to the pixel coordinate system of the high-resolution color camera is determined. In addition, this paper designs experiments to verify the performance of the model. The research results show that the method proposed in this paper has a certain effect.