The measurement of a fatigue crack propagation path is very important to the study of fracture mechanics. For the in-plane biaxial crack propagation test, it is more difficult to accurately measure the crack path due to the unique bifurcation and turning phenomenon of the crack. This study provides an in situ biaxial crack growth path automatic tracking measurement system and method based on image recognition technology. The system includes the in situ biaxial mechanical test system, an optical microscope and intelligent camera, a microscope-fixed bracket, and the XY electric mobile platform of the testing machine. During the test, the optical microscope and intelligent camera were fixed, and the crack tip morphology was automatically identified by an artificial intelligence crack identification and tracking system based on the convolutional neural network (CNN) developed by Python language. The camera transmits the coordinates of the crack tip to the XY platform control software in real time, and it automatically adjusts the position of the XY moving platform that drives the loading system to move, ensuring that the crack tip is always in the center of the field of view during the crack propagation process. The system records the crack tip coordinates and cycle times according to the set requirements to fully track the crack growth process. The system can record crack tip coordinates and monitor the crack growth state, which can improve the efficiency and measurement accuracy of crack propagation tests.