Non-intrusive load identification based on characteristic values such as voltage, current, and power exhibits a certain degree of fuzziness and randomness. In complex scenarios, traditional methods often struggle to accurately identify load types due to insufficient feature extraction, leaving room for improvement in accuracy. To address this issue, this paper proposes a non-intrusive residential load identification method based on the MTF (Markov Transition Field)-Vit (Vision Transformer) algorithm. The Markov Transition Field is used to transform the current, voltage, and frequency data of six types of loads into higher-dimensional RGB images. The generated RGB images retain all the features of the original data, and the Vision Transformer image classification model is employed for feature extraction and classification. First, the load types are encoded and preprocessed. Then, an MTF-Vit classification model is constructed, and an image classification model is iteratively generated after inputting the data. Finally, case studies are conducted to validate the proposed method. Experimental comparisons with VGG (Visual Geometry Group), ResNet (Residual Network), and CNN (Convolutional Neural Network) models demonstrate that the model utilizing the Vision Transformer algorithm significantly improves convergence speed and fitting accuracy, achieving an accuracy of 96%. The proposed method yields promising results in non-intrusive load identification and provides new insights and approaches for this field.