Solar cell surface defect detection is an indispensable process in the production of photovoltaic modules. Automatic defect detection methods based on machine vision are widely used due to their high accuracy， real-time and low cost advantages. This paper reviewed the research progress of machine vision-based solar cell surface defect detection methods. First， the solar cell surface imaging method was described and typical defect types were listed. Then， the principles of solar cell surface defect detection based on traditional machine vision algorithms and based on deep learning algorithms were analyzed， respectively. The traditional machine vision algorithms were reviewed in terms of image domain analysis， transform domain analysis； the research status of solar cell surface defect detection based on deep learning in recent years was outlined in terms of unsupervised learning， supervised learning and weakly supervised and semi-supervised learning， respectively. Various typical methods for solar cell surface defect detection were further subdivided into categories and comparative analysis， and the advantages and disadvantages of each method were summarized. Subsequently， nine types of solar cell surface defect image datasets and defect detection performance evaluation metrics were introduced. Finally， the common key problems of solar cell defect detection and their solutions were summarized systematically， and the future development trend of solar cell surface defect detection was foreseen.