As the substrate of electronic devices, the printed circuit board (PCB) is in high demand. It carries the layout of circuit components and wires, which has a significant impact on the quality of electronic products. Because electronic devices are thin and compact, PCB defect detection using machine vision is difficult. This article examines PCB defect detection algorithms based on machine vision in recent 10 years from three perspectives, including classical image processing, traditional machine learning, and deep learning. The advantages and disadvantages are analyzed to improve researchers′ understanding of PCB defect detection. Nine PCB data sets and performance evaluation indexes are introduced. The advanced algorithms are compared and analyzed on PCB data sets and popular small target data sets, respectively. Finally, the current challenges with the PCB defect detection method are discussed and future research objectives.