Feasibility discrimination of anchor-relieving robot anchor-relieving operation based on anchor cable state image recognition

Abstract Considering that the dispersion and bending state of the anchor cable will affect the breaking operation of the anchor cable cutting tool to the anchor cable, this paper studies the anchor cable state recognition method based on image recognition. Firstly, the YOLOv8 algorithm is used to distinguish the dispersion state of the anchor cable, and then the semantic segmentation and image processing methods are used to identify the bending state of the anchor cable.The average detection accuracy of the anchor cable dispersion state recognition method based on YOLOv8 algorithm for the five types of targets of intact anchor cable, slight dispersion and complete dispersion, and locks and trays reached 89.4%.The detection accuracy of the three states of intact anchor cable, slight dispersion and complete dispersion reached more than 90%, indicating that the detection model has a good recognition effect of anchor cable dispersion state.The bending state discrimination method of anchor cable based on semantic segmentation and image processing method can realize the bending angle discrimination of intact anchor cable, overall bending and partial bending anchor cable.Firstly, the YOLOv8 target detection model is used to detect the target position of the anchor cable and the lock, and then the target is framed.Then the image in the target box is cut off, and the target segmentation of the lock and anchor cable is performed by the PSPnet semantic segmentation model.Then, the segmented mask map is transformed into an hsv color gamut, and then the lock and anchor cable are screened out by red threshold and green threshold respectively.Then, the contours of the lock and the anchor cable are detected respectively, and the minimum circumscribed rectangle of the contour is calculated.Finally, the midline of the minimum outer rectangular long side of the lock and the anchor cable is calculated respectively, and the angle between the two midlines is calculated, that is, the bending angle of the anchor cable.When the camera is perpendicular to the anchor plane, the calculated angle error is less than 2°.Through the identification of the dispersion and bending state of the anchor cable, it can be judged whether the anchor withdrawal device can complete the breaking operation of the anchor cable.