Transfer Learning-Based Image Processing of Recognizing Office Objects for Indoor Mobile Robots
Ze Gong, Longfei Su, Hongshuo Liu, Yao Gu, Ming Yue · 2023
This paper proposes an OpenCV-based image processing method and a target recognition approach for indoor office mobile robots based on transfer learning. Firstly, the recognized image is pre-processed for clarity to remove the effect of uneven lighting; secondly, the transfer learning idea is introduced to extract image features, then the study modifies the fully connected layer of the Resnet18 network and trains the network model parameters; finally, the trained network model is used to conduct target recognition experiments on the pre-processed image. The experimental results show that the method can effectively recognize the images of items in indoor office mobile robots with relatively high accuracy and efficiency.