The invention relates to an adaptive binocular vision turbine blade detection device. The operating principle of the device comprises the following steps: performing image sampling on a detected turbine blade through a binocular vision system, extracting characteristics of a specified position point of the turbine blade, rapidly identifying a model of the detected turbine blade by an identification algorithm in a microprocessor, and finally, controlling a servo motor to automatically adjust space positions of two charge coupled device (CCD) cameras relative to the detected turbine blade by the microprocessor. Finally, the aim of rapidly and accurately detecting turbine blades of various models is achieved.