Research and Application of the Obstacle Avoidance System for High-Speed Railway Tunnel Lining Inspection Train Based on Integrated 3D LiDAR and 2D Camera Machine Vision Technology

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(13), P. 7689 - 7689

Published: June 29, 2023

This study presents an innovative, intelligent obstacle avoidance module intended to significantly enhance the collision prevention capabilities of robotic arm mechanism onboard a high-speed rail tunnel lining inspection train. The proposed employs fusion ORB-SLAM3 and Normal Distribution Transform (NDT) point cloud registration techniques achieve real-time densification, ensuring reliable detection small-volume targets. By leveraging spatial filtering, cluster computation, feature extraction, precise localization information is further obtained. A multi-modal data achieved by jointly calibrating 3D LiDAR camera images. Upon validation through field testing, it demonstrated that can effectively detect obstacles with minimum diameter 0.5 cm, average deviation controlled within 1–2 cm range safety margin 3 preventing collisions. Compared traditional sensors, this provides across more dimensions, offering robust support for construction powerful automated control systems digital twin lifecycle analysis railway tunnels.

Language: Английский

---

Enhanced Hyperspectral Image Classification Through Pretrained CNN Model for Robust Spatial Feature Extraction

Journal of Optics, Journal Year: 2023, Volume and Issue: 53(3), P. 2287 - 2300

Published: Nov. 18, 2023

Language: Английский

---

Strategic analysis of intelligent connected vehicle industry competitiveness: a comprehensive evaluation system integrating rough set theory and projection pursuit

Complex & Intelligent Systems, Journal Year: 2024, Volume and Issue: 10(5), P. 7033 - 7062

Published: June 29, 2024

Abstract As a carrier of multi-industrial technology integration and the key to industrial competition, intelligent connected vehicle (ICV) has been taken seriously around world. However, as fast-growing emerging industry, its development process varies greatly from place place. Hence, merits demerits are analyzed for ICV industry in different cities scientifically clarify links each city, this paper suggests an extensive assessment framework integrating rough set theory projection pursuit-based computation systematically assess thoroughly evaluate level competitiveness industry. First, through big data text analysis technology, we constructed "5 + 24" two-tier evaluation index system composed 24 level-II indexes well five level-I selected 19 typical input comprehensive system. Further, Adaptive Random Forest based Crossover Tactical Unit (ARF-CTU) algorithm is proposed evaluating performance ARF employed estimate lowering overfitting issues handling high dimensional data. Moreover, continuously varying conditions by CTU. Then, pursuit: (I) Quoting non-decision-making attribute reduction, that is, under premise unchanged classification ability, derive new system, calculate weight score on (II) Based pursuit mapped genetic linear structure, one-dimensional vector output.

Language: Английский

---

A Nonlinear Convolutional Neural Network Algorithm for Autonomous Vehicle Lane Line Detection

Scalable Computing Practice and Experience, Journal Year: 2024, Volume and Issue: 25(5), P. 4237 - 4245

Published: Aug. 1, 2024

The traditional lane line detection algorithm relies on artificial design features, which has poor robustness and cannot cope with the complex urban street background. With rise of deep learning technology, model convolutional neural network as mainstream is widely used in field computer vision, provides a new idea for detection. In order to improve disadvantages methods that are vulnerable environmental impact robustness, nonlinear convolution driverless proposed. Firstly, pretreatment method extracting region interest enhancing contrast lines reduce unnecessary image background enhance feature details image. Existing learning-based algorithms still have difficulties. First, accumulated wear tear will cause fade fade; roadside trees buildings can interfere performance algorithm. addition, compared pixels whole picture, too few, layer easy lead loss details. when traffic flow large, easily blocked, makes it more difficult detect line. Then built based features extracted by CNN, DBSCAN clustering post-process segmentation model; Finally, least square fit quadratic curve pixel peak points line, fitting results regressed original experimental show accuracy recall verification set 91.3% 90.6%, respectively, indicating good effect. It proved CNN combined post-processing effectively defects experience, better than method.

Language: Английский

---

Stacked ensemble model for analyzing mental health disorder from social media data

Multimedia Tools and Applications, Journal Year: 2023, Volume and Issue: 83(18), P. 53923 - 53948

Published: Nov. 27, 2023

Language: Английский

---