A Utility Pole Detection Model Based on YOLOv8s

Traditional utility pole inspection methods primarily rely on manual labor, which is not only inefficient but also susceptible to weather, environmental, and human factors, leading to potential hazards being undetected and unaddressed in a timely manner. This paper proposes a utility pole detection model named CGN-YOLOV8s ,which is based on YOLOv8s. The model is built upon the Depthwise Separable Convolution-Convolutional Block Attention Module (DSConv-C2f-CBAM), which independently computes the spatial feature relationships and inter-channel feature relationships of images, thereby enhancing the detection capability for small target objects and effectively addressing the difficulty of detecting small targets in utility pole detection tasks. Simultaneously, by integrating the Depthwise Separable Convolution-Global Attention Module (DSConv-SPPF-GAM), the model leverages the global attention mechanism to strengthen feature representation and cross-dimensional interaction capabilities, solving the detection problem when utility poles are obscured. Furthermore, by designing the NWD-CIOU loss function, the model further improves the localization accuracy of utility pole targets in complex environments and reduces missed detections. This model demonstrates high efficiency and accuracy in utility pole detection tasks, meeting the demands of the power industry for utility pole detection and providing robust support for the safe operation of power systems