STIF: A Spatial–Temporal Integrated Framework for End-to-End Micro-UAV Trajectory Tracking and Prediction With 4-D MIMO Radar

IEEE Internet of Things Journal, Journal Year: 2023, Volume and Issue: 10(21), P. 18821 - 18836

Published: Feb. 16, 2023

The early trajectory prediction of micro unmanned aerial vehicles (micro-UAVs) with random behavior intentions facilitates the elimination potential safety hazards. However, due to property a small radar cross Section (RCS), backscattered signals from micro-UAVs may be submerged under strong background clutters, leading distorted tracking and false prediction. To this end, article presents spatial–temporal integrated framework (STIF) for end-to-end micro-UAV based on 4-D multiple-input–multiple-output (MIMO) radar. Especially, obtain accurate trajectories in low signal-to-noise ratio (SNR) conditions, target detection are considered interdependent addressed jointly work, rather than treating them as two separate processes conventional methods. advantage is that assistance tracking, all consecutive spatial information encoded raw streams can incorporated enhance continuous performance, avoiding loss using only one single scan. Subsequently, accommodate high maneuvering scenarios, an intention-aware transformer-based presented simultaneously discover both temporal dependencies hiding long-term estimated trajectories. Consequently, frequency modulated wave (FMCW) utilized evaluate proposed system. Numerous simulation experimental results indicate STIF outperforms competing state-of-the-art methods achieve superior performance accuracy 0.3851 m SNR conditions.

Language: Английский

---

Vision-Based Robust Lane Detection and Tracking in Challenging Conditions

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 67938 - 67955

Published: Jan. 1, 2023

Lane marking detection is fundamental for both advanced driving assistance systems and traffic surveillance systems. However, detecting lane highly challenging when the visibility of a road low, obscured or often invisible due to real-life environment adverse weather. Most methods suffer from four types challenges: (i) light effects i.e. shadow, glare light, reflection etc. created by different sources like streetlamp, tunnel-light, sun, wet etc.; (ii) Obscured eroded, blurred, dashed, colored cracked caused natural disasters weather (rain, snow etc.); (iii) occlusion objects surroundings (wiper, vehicles (iv) presence confusing lines inside view e.g., guardrails, pavement marking, divider In this paper, we proposed simple, real-time, robust tracking method detect considering abovementioned conditions. method, introduced three key technologies. First, introduce comprehensive intensity threshold range (CITR) improve performance canny operator in edges clear, low intensity, cracked, colored, blurred edges. Second, propose two-step verification technique, angle-based geometric constraint (AGC) length-based (LGC) followed Hough Transform, verify characteristics prevent incorrect detection. Finally, novel predict position next frame defining horizontal (RHLP) along x axis which will be updating with respect previous frame. It can keep track either left right markings are partially fully invisible. To evaluate used DSDLDE [1] SLD [2] dataset 1080 ×1920 480×720 resolutions at 24 25 frames/sec respectively where video frames containing scenarios. Experimental results show that average rate 97.55%, processing time 22.33 msec/frame, outperform state-of-the-art method.

Language: Английский

---

Fusion of Deep Sort and Yolov5 for Effective Vehicle Detection and Tracking Scheme in Real-Time Traffic Management Sustainable System

Sustainability, Journal Year: 2023, Volume and Issue: 15(24), P. 16869 - 16869

Published: Dec. 15, 2023

In recent years, advancements in sustainable intelligent transportation have emphasized the significance of vehicle detection and tracking for real-time traffic flow management on highways. However, performance existing methods based deep learning is still a big challenge due to different sizes vehicles, occlusions, other scenarios. To address issues, an effective scheme proposed which detects vehicles by You Only Look Once (YOLOv5) with speed 140 FPS, then, Deep Simple Online Real-time Tracking (Deep SORT) integrated into result track predict position vehicles. first phase, YOLOv5 extracts bounding box target second it fed output perform tracking. Additionally, Kalman filter Hungarian algorithm are employed anticipate final trajectory evaluate effectiveness algorithm, simulations were carried out BDD100K PASCAL datasets. The surpasses learning-based methods, yielding superior results. Finally, multi-vehicle process illustrated that precision, recall, mAP 91.25%, 93.52%, 92.18% videos, respectively.

Language: Английский

---

Enhancing Autonomous Vehicle Navigation Through Computer Vision: Techniques for Lane Marker Detection and Rain Removal

Lecture notes in intelligent transportation and infrastructure, Journal Year: 2025, Volume and Issue: unknown, P. 167 - 191

Published: Jan. 1, 2025

Language: Английский

---

Exploring the Impact of Deep Learning Models on Lane Detection Through Semantic Segmentation

SN Computer Science, Journal Year: 2024, Volume and Issue: 5(1)

Published: Jan. 3, 2024

Language: Английский

---

Computer Vision-Based Lane Detection and Detection of Vehicle, Traffic Sign, Pedestrian Using YOLOv5

Sakarya University Journal of Science, Journal Year: 2024, Volume and Issue: 28(2), P. 418 - 430

Published: April 26, 2024

There has been a global increase in the number of vehicles use, resulting higher occurrence traffic accidents. Advancements computer vision and deep learning enable to independently perceive navigate their environment, making decisions that enhance road safety reduce Worldwide accidents can be prevented both driver-operated autonomous by detecting living inanimate objects such as vehicles, pedestrians, animals, signs well identifying lanes obstacles. In our proposed system, images are captured using camera positioned behind front windshield vehicle. Computer techniques employed detect straight or curved images. The right left within driving area vehicle identified, drivable is highlighted with different color. To signs, cars, bicycles around vehicle, we utilize YOLOv5 model, which based on Convolutional Neural Networks. We use combination study-specific GRAZ dataset research. object detection study, involves 10 objects, evaluate performance five versions model. Our evaluation metrics include precision, recall, precision-recall curves, F1 score, mean average precision. experimental results clearly demonstrate effectiveness lane method.

Language: Английский

---

Probabilistic Hough Transform for Rectifying Industrial Nameplate Images: A Novel Strategy for Improved Text Detection and Precision in Difficult Environments

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(7), P. 4533 - 4533

Published: April 3, 2023

Industrial nameplates serve as a means of conveying critical information and parameters. In this work, we propose novel approach for rectifying industrial nameplate pictures utilizing Probabilistic Hough Transform. Our method effectively corrects distortions clipping, features collection challenging analysis. To determine the corners nameplate, employ progressive Probability Transform, which not only enhances detection accuracy but also possesses ability to handle complex scenarios. The results our are clear readable text, demonstrated through experiments that show improved in model identification compared other methods.

Language: Английский

---

Lane Detection in Autonomous Driving: A Comprehensive Survey of Methods and Performance

Published: Jan. 8, 2024

Language: Английский

---

Probabilistic Hough Transform for Rectifying Industrial Nameplate Images: A Novel Strategy for Improved Text Detection and Precision in Difficult Environments

Published: March 17, 2023

Industrial nameplates serve as a means of conveying critical information and parameters. In this work, we propose novel approach for rectifying industrial nameplate pictures utilizing probabilistic Hough transform. Our method effectively corrects distortions clipping, features collection challenging analysis. To determine the corners nameplate, employ progressive probability transform, which not only enhances detection accuracy but also possesses ability to handle complex scenarios. The results our are clear readable text, demonstrated through experiments that show improved in model identification compared other methods.

Language: Английский

---

A COMPREHENSIVE APPROACH TO AUTONOMOUS VEHICLE NAVIGATION

Journal of Al-Azhar University Engineering Sector, Journal Year: 2024, Volume and Issue: 0(0), P. 167 - 182

Published: July 29, 2024

Autonomous vehicles are revolutionizing transportation, and the accuracy of road lane detection is a pivotal aspect this innovation. This paper presents an in-depth exploration sophisticated system, geometric modeling to estimate structure boundaries based on images captured by onboard vehicle camera, deployment object techniques. The system meticulously designed, employing series computer vision techniques identify track lanes in various driving conditions. curve fitting component utilizes second-order polynomial, providing mathematical model that accurately represents curvature intricate dynamics detected lanes. representation provides more nuanced understanding geometry, aiding prediction trajectory. facet research focuses recognition classification objects within environment, contributing significantly overall situational awareness autonomous systems. YOLO (You Only Look Once) algorithm commonly used for purpose as it can process frames at impressive speed while maintaining high accuracy, making suitable real-time applications. efficacy suggested was confirmed conducting experiments two distinct datasets. proposed method achieved 98.64% Tusimple 96.92% KITTI dataset, demonstrating its robustness reliability under varying Special Issue AEIC 2024 (Electrical System & Computer Engineering Session)

Language: Английский

---