Within the theoretical framework of 3D point cloud data, this paper proposes the use of laser radar sensors to collect 3D point cloud data of cable size features. Due to the presence of redundant interference data in the data, an adaptive filtering algorithm is used to preprocess the data. To better extract cable dimension features, a cable dimension feature extraction model based on the ADGCNN network is designed. Through feature enhancement and fusion, a deep learning training model for cable dimension features is established. To address the issue of suboptimal model training performance, the Adadelta optimization algorithm is applied to optimize the model, and its optimization effects are verified and analyzed. The accuracy rate before model optimization was 0.894. After applying the Adadelta optimization algorithm, the model’s accuracy rate improved to 0.975, confirming the effectiveness of the Adadelta optimization algorithm in model optimization.
