The application discloses a
crusher material automatic grading control method and
system based on image recognition, which comprises the following steps: collecting a dynamic
image sequence, extracting deformation gradient, normal change rate and edge disturbance features; collecting torque signals and the like, constructing a dynamic sequence and extracting a
chaotic dynamic feature vector; collecting stress and
acoustic vibration signals, and fusing to generate a multi-field collaborative
feature vector; inputting an improved ResNeXt network, fusing three types of features, and outputting particle size classification and deviation information; constructing a grading evolution controller, performing strategy adaptive reconstruction, and generating
optimal control instructions; and executing
discharge port, rotating speed and feeding adjustment, and closing loop feedback to the previous process to realize
dynamic control. Through the fusion of dynamic image topological features, crushing dynamics features and multi-
physical field features, and in combination with an
adaptive evolution control mechanism, high-precision identification and full-automatic optimal adjustment of the particle size of the
crusher material are realized.