Subgrade two-side visual imaging monitoring invader situation identification method

The invention discloses a roadbed two-side visual imaging monitoring invading object situation identification method, which comprises the steps of 1, dividing a training image set and a verification image set, 2, establishing an invading object situation identification pre-judgment model and training the model, 3, verifying and optimizing the invading object situation identification pre-judgment model, and 4, identifying the invading object situation according to the invading object situation identification pre-judgment model. Step 4, obtaining and processing a dynamic posture image set of the to-be-detected area; step 5, outputting a dynamic posture recognition pre-judgment result of the to-be-detected area by the model; according to the invention, through the gesture recognition pre-judgment model established based on the deep learning neural network and the pre-filtering gradient method, real-time recognition and pre-judgment of the next motion of the gesture can be carried out on the dynamic gesture of th