Misalignment Detection Algorithm for Vertical Rigid Tank Channel Joints Based on Improved Yolov8n

In order to discover the faults of vertical well rigid tank channel joints in time and eliminate the hidden dangers of hoisting system operation, a YOLOv8n misalignment detection algorithm for vertical well rigid tank channel joints, with the improvement of YOLOv8n, is proposed as YOLOv8n-API. Initially, the YOLOv8n central network has been upgraded by integrating the CA attention mechanism module. This integration primarily aims to counteract the disturbances caused by the inconsistency and faintness of feature details in images captured inside the wellbore. It achieves this by merging channel dimensions with spatial relational position data, thereby boosting the network’s ability to extract finer feature details. Subsequently, to diminish the computational load, the C2f module’s complexity is lessened by replacing Bottleneck with the more efficient Faster_Block. Furthermore, to tackle the adverse effects of poor-quality images, the WIOU loss function is utilized in place of CIOU, enhancing the precision of anchor frame prediction. Lastly, a variable focus ranging technique is employed for determining the offset dimensions of tank channel joints. Test results indicate that, in comparison with the standard YOLOv8n network, the YOLOv8n-API detection network exhibits a 6% enhancement in mAP and achieves a reduction of 1.4 MB in model size. These improvements were observed on the validation set for rigid tank channel joints in vertical wells, and compared with other YOLO algorithms, YOLOv8n-API has a significant advantage in the detection of misalignment of joints of vertical wells with rigid tank channels.