Working stage identification of hydraulic excavator

The identification of the working stage is of great significance in the study of energy saving of excavators. In this paper, computer vision is introduced into the excavator's work cycle stage recognition, and the YoloV2 algorithm is used to establish a deep vision target detector to locate and classify these three characteristic parts (bucket, the arm joint and the body). In order to solve the problem of visual misjudgment, this paper proposes a method that comprehensively considers pump pressure information and visual information. Combining the general action sequence of the excavator under normal operation, a state machine is established. All working stages are set to states, and transition conditions are set according to pressure and visual information to realize state switching. This method has been applied in the working process of a certain model of excavator, and the various working stages can be correctly judged according to the operating results.