Application of an Improved YOLOv5 Algorithm in Real-Time Detection of Foreign Objects by Ground Penetrating Radar

Abstract

Ground penetrating radar (GPR) detection is a popular technology in civil engineering. Because of its advantages of non-destructive testing (NDT) and high work efficiency, GPR is widely used to detect hard foreign objects in soil. However, the interpretation of GPR images relies heavily on the work experience of researchers, which may lead to problems of low detection efficiency and a high false recognition rate. Therefore, this paper proposes a real-time detection technology of GPR based on deep learning for the application of soil foreign object detection. In this study, the GPR image signal is obtained in real time by the GPR instrument and software, and the image signals are preprocessed to improve the signal-to-noise ratio of the GPR image signals and improve the image quality. Then, in view of the problem that YOLOv5 poorly detects small targets, this study improves the problems of false detection and missed detection in real-time GPR detection by improving the network structure of YOLOv5, adding an attention mechanism, data enhancement, and other means. Finally, by establishing a regression equation for the position information of the ground penetrating radar, the precise localization of the foreign matter in the underground soil is realized.

An Analysis of Defect Detection on Steel Strip Surface Using Modified Yolo V5

“…The 2020 YOLOv5 proposal by Glenn Jocher is based on the cutting-edge optimization methods used in the YOLO object detection architecture. Even though, there are no relevant papers published of YOLOv5, it has been accepted widely and the code is still being updated [31]. YOLOv5, which is 90 percent smaller than YOLOv4 and faster than earlier iterations, The PyTorch framework, on which YOLOv5 is based, offers conversion to ONNX and CoreML, which makes it simple to deploy on embedded and mobile devices.…”

“…(4)HEAD: In order to anticipate class and image features and create bounding boxes around the target item, the head or output part of YOLOv5 is used. The head section outputs through target predictions that are made [31].…”

“…The IOU score is based on how accurately the target's bounding box is placed [29]. IOU calculates this by dividing the side lengths of the intersection of the detection box and the ground-truth box by the side lengths of the union [31].Correct and incorrect detections are provided by IOU threshold. The IOU and IOU loss function are defined as follows.…”

Improving FMCW GPR Precision through the CZT Algorithm for Pavement Thickness Measurements

“…The traditional GPR system is an impulse radar, which works by sending electromagnetic (EM) pulses through an antenna to the road surface and then recording the reflected pulses from the internal interface. The authors of [5][6][7][8][9][10][11][12][13][14][15][16] conducted in-depth research on pavement thickness using GPR, and the thickness information of the road surface is determined by the pulse time. Since the pavement thickness is mostly less than 10 cm, the pulse peak difference is often only a few nanoseconds.…”

Mining belt foreign body detection method based on YOLOv4_GECA model

“…Among the results, the fastest detection speed was obtained with the combination of SSD and MobileNet.In 2020, Pang 15 et al proposed a real-time detection method based on the YOLOv3 algorithm to detect hidden metal weapons on the human body, which was applied to passive millimeter wave (PMMW) images. It not only has high accuracy but also very fast detection speed in terms of detection for targets with small size.In 2022, Chen 16 et al combined the YOLOv4 algorithm with an optimized anchor box to achieve efficient detection of foreign objects in belt conveyors and reduce the occurrence of this problem of longitudinal belt tears.In 2022, Qiu 17 et al A deep learning-based real-time detection technique for ground radar with added attention mechanism and data augmentation to improve false and missed detection problems in detection.In 2022, Jing 18 et al proposed a random forest framework based on optimal pixel vision features and designed pixel vision features (PVF) in order to overcome the complexity of airport pavement image information and the variability of foreign object fragments, which is more advantageous in terms of accuracy and recall of foreign object fragment detection.In 2022, Abramson 19 et al created a fully automated foreign object tracking algorithm that used a custom convolutional neural network to achieve 99% accuracy, surpassing other comparable algorithms.…”