Urban Underground Target Detection from Ground Penetrating Radar Images Using YOLOX
Ground penetrating radar(GPR)has been widely used in urban road detection due to its non-destructive,high efficiency,and high resolution characteristics.However,its data interpretation mainly relies on experts'experience,which is time-consuming and laborious,and will lead to a large rate of misjudgment and missed judgment.An automatic urban underground target detection method based on YOLOX is proposed to realize the automatic interpretation of GPR data.This method uses the YOLOX network,which has the ability of simultaneously automatic feature extraction and automatic target location,to realize the detection of typical urban underground targets(cavities,cables,and pipes)in GPR images.Since the amount of GPR data is small,the network is first pre-trained on the COCO public dataset,and then the parameters are fine-tuned on the GPR dataset collected in the indoor laboratory by using transfer learning.Fi-nally,the trained YOLOX network is used to locate and identify underground targets.Experimental results show that the proposed method can accurately detect urban underground targets under small samples.The mean average precision(mAP(0.75))can reach 82.1%and it can realize real-time detection.Moreover,the proposed method outperforms the state-of-the-art methods in terms of detection precision.
万方数据
维普
