Application of target recognition technology based on visual image processing in mobile robot

DOI:	10.11809/bqzbgcxb2024.03.041
Keywords:	human tracking; face recognition; image processing; Faster RCNN; Kalman filter; Facenet network
Abstract:	In order to improve the recognition performance of moving human bodies in mobile robots, the study utilizes Faster RCNN KF for real time tracking of dynamic human body images, and combines Facenet MTCNN to achieve facial recognition of tracked objects. The results of human motion tracking and detection experiments and facial recognition tests show that the tracking error of the Faster RCNN KF algorithm proposed in the study is only 0.000 5 m, and the tracking response speed and error correction speed are fast. At the same time, the Facenet MTCNN target recognition algorithm proposed in the study has a classification accuracy of up to 99.15% in training, and a time delay of 0.01 seconds in classification, which can effectively identify the identity information of the tracked object. The research results indicate that visual image processing technology can achieve effective tracking and detection of the human body, and can recognize faces of tracked objects with different identities, which is of great value for the development of mobile robot tracking and recognition technology.
Published:	2024-03-28
PDF