The invention belongs to the technical field of AR (augmented reality) and provides an opencv and easyar based virtual key trigger method. The method includes: S1, calibration, to be more specific, presetting a target with a calibration object, and acquiring picture information in real time; S2, sampling, to be more specific, when the calibration object is searched, acquiring misjudgment areas, and recording quantity of misjudgment areas passing through a virtual key position within sampling time; S3, tracking recognition, to be more specific, tracking the calibration object in real time, counting the misjudgment areas passing through a preset virtual key position within each sampling time, and when the quantity is larger than an initial number, running virtual key associated operations. By means of compromise processing, existing high-efficiency training files are adopted for hand training, training results are subjected to secondary operation, and accordingly hand position recognition can be realized, producitng standards of training files high in accuracy requirement are lowered, and low operation quantity and quickness are realized.