Rail surface damage detection method based on depth vision algorithm
[Objective]The expansion of railroad operations has facilitated the research on the timely and accurate detection of rail surface injuries.Traditional methods of rail surface defect detection heavily relied on manual inspections.Although these methods offer benefits such as simple operation and low cost,they possess drawbacks such as low efficiency,high rates of missed defects,and poor real-time performance.Automated detection techniques such as ultrasonic,eddy current,and magnetic flux leakage detection are significant in rail damage detection.However,they possess limitations such as high dependence on hardware and the need for skylight point operation,leaving some rail injuries unaddressed in a timely manner.Furthermore,despite the advances in existing deep learning-based algorithms for rail defect detection,they face issues such as inference delays,instability,and high computational costs,failing to meet the real-time requirements of field operations.To overcome these limitations,we propose an innovative algorithm based on deep learning theory,referred to as RT-DETR,to achieve efficient and precise detection of rail surface injuries.[Methods]The RT-DETR algorithm eliminates the need for non-maximum suppression(NMS)in traditional target detection algorithms,which helps avoid the inference delays and instability associated with existing detectors.Based on the single-scale feature interaction module(AIFI)with the attention mechanism and the cross-scale feature fusion module(CCFM)based on CNN,we introduce an efficient hybrid encoder that decouples single-scale internal interactions and cross-scale fusion,replacing the traditional transformer encoder.An IoU-aware initialized object query mechanism is proposed to enhance the initialization of decoder queries.The objective function is redefined to better balance classification scores with IOU scores and improve detection performance.This further ensures consistency in the classification and localization of positive-sample targets,thus enabling the model to produce higher-quality encoder features.This experimental scheme has helped develop a real-time target detector that is applicable in real-time industrial inspection.[Results]The RT-DETR-L model was selected for the algorithm experiment,with comparative tests conducted against the DETR and YOLOv8l algorithms.The experimental results are summarized below.1)At iteration 299,the RT-DETR-L model exhibited a total loss of 1.222,which included a classification loss of 0.155,a localization loss of 0.798,and a confidence loss of 0.269.The RT-DETR-L model exhibited a significantly low loss value and high iteration speed.2)In the comparative experiment,the RT-DETR-L algorithm achieved a peak mean average precision(mAP)of 96.167% at the 265th iteration,outperforming both the DETR and YOLOv8l algorithms in terms of detection accuracy.3)The RT-DETR-L program significantly enhanced the precision and recall rates for detecting rail surface injuries.Further,it effectively identified various types of defects,including detecting spalling,head check,and rail gap,yielding accuracies of 95.1%,93.8%,and 99.5%,respectively.Additionally,it achieved a detection speed of 8.62 ms/frame with a minimal resource footprint,requiring only 4.2 M parameters.[Conclusions]Compared to existing mainstream algorithms in this field,the proposed RT-DETR-L algorithm demonstrates faster inference speeds and higher detection accuracy for real-time rail injury detection.These research outcomes indicate that we have realized a reliable and efficient detection solution for future rail maintenance and repairs,and thus,we have effectively addressed the problems of delayed inference and high computational demands associated with traditional rail injury detection methods.Additionally,they indicate that we have compensated for the deficiencies in conventional target detection approaches.
万方数据
维普
