The rheological property of paste is the foundation of the paste-filling process in metal mines and an important index in paste-filling technology. In this paper, a method of predicting the paste yield stress using machine vision is proposed through the development of image acquisition standards and an image acquisition device to collect image data sets based on a paste image data set. The 3D convolutional neural networks (3D CNNs) network model is used to extract the depth features and timing information on the paste. An image enhancement strategy for the histogram equalization algorithm is introduced to reduce the interference of environmental factors, and the yield stress value corresponds to the image information. The preprocessed data set is used for training and testing the 3D CNNs network model. Additionally, the 3D CNNs network model is optimized, and the prediction accuracy of the model is increased from 93.47% to 98.54%, demonstrating the feasibility of paste yield stress prediction based on machine vision. Compared with the traditional paste rheological measurement method, it solves the problems of complex operation of traditional paste yield stress measurement, strong interference of external factors, and the difficulties associated with engineering sites.