The invention discloses a process model graph automatic identification and understanding method, and belongs to the field of process mining. The method comprises the steps of firstly constructing a basic graphic element template, matching a process model graph with the graphic element template, identifying model elements such as tasks, activities, events, gateways, arrowheads and the like in the process model graph, and removing repeated matching nodes and wrong matching regions by using a screening technology; secondly by using a process model graph cutting technology, obtaining pictures of regions where nodes comprising texts are located, and by using an OCR technology, recognizing the texts in the pictures; thirdly performing gray processing on the process model graph, and obtaining andstoring a gray value matrix of the process model graph; and finally according to the arrowheads in the model graph and the positions of nearest neighbor nodes, performing traversal in the gray valuematrix, and identifying a starting node and a terminating node of a directed edge. According to the method, the types and positions of the model nodes and the texts in the model nodes can be correctlyidentified, and also the directed edge in the process model graph can be correctly identified.