A Systematic Review on Vision‐Based Traffic Density Estimation for Intelligent Transportation Systems DOI Creative Commons

Muhammad Ardi Putra,

Agus Harjoko, Wahyono Wahyono

et al.

IET Intelligent Transport Systems, Journal Year: 2025, Volume and Issue: 19(1)

Published: Jan. 1, 2025

ABSTRACT Traffic congestion is often considered one of the major challenges faced in urban areas. It important to address this issue due its significant negative impacts on both society and environment, including decreased productivity increased pollution. For reason, implementing a traffic density estimation system necessary as it can be further integrated into adaptive control systems that dynamically adjust lights based real‐time levels. Different from existing papers categorise vision‐based methods microscopic macroscopic approaches, paper contributes novel taxonomy by introducing hybrid approach, which combines two leverage their respective advantages. Furthermore, review offers guidance for future research topic. Later discussion, three approaches estimating will broken down specific used, namely image processing techniques, machine learning models, deep or combination them. This also provides coherent discussion details these papers, well advantages drawbacks. To best our knowledge, first specifically discusses exclusively video data.

Language: Английский

Enhancing Retrieval-Oriented Twin-Tower Models with Advanced Interaction and Ranking-Optimized Loss Functions DOI Open Access
Ganglong Duan, Shanshan Xie, Yutong Du

et al.

Electronics, Journal Year: 2025, Volume and Issue: 14(9), P. 1796 - 1796

Published: April 28, 2025

This paper presents an optimized twin-tower model for text retrieval that addresses limitations in traditional models through improved feature interaction and loss function design. We introduce early layer using cross-attention mechanisms a ranking-optimized function. These innovations enable earlier interactions between queries documents, enhance semantic relationship understanding, optimize relative similarity rankings while reducing overfitting risk. Our experiments on NQ, TQA, WQ datasets show substantial Top-K accuracy improvements over benchmark like BM25, DPR, ANCE, ColBERT. For example, our achieves 20.3% improvement Top-20 NQ compared to with only 17 ms latency. Ablation studies confirm the effectiveness of improvements. research demonstrates enhancing optimizing functions significantly improves performance, providing valuable methodological insights efficient maintaining computational efficiency.

Language: Английский

Citations

0
A Systematic Review on Vision‐Based Traffic Density Estimation for Intelligent Transportation Systems DOI Creative Commons

Muhammad Ardi Putra,

Agus Harjoko, Wahyono Wahyono

et al.

IET Intelligent Transport Systems, Journal Year: 2025, Volume and Issue: 19(1)

Published: Jan. 1, 2025

ABSTRACT Traffic congestion is often considered one of the major challenges faced in urban areas. It important to address this issue due its significant negative impacts on both society and environment, including decreased productivity increased pollution. For reason, implementing a traffic density estimation system necessary as it can be further integrated into adaptive control systems that dynamically adjust lights based real‐time levels. Different from existing papers categorise vision‐based methods microscopic macroscopic approaches, paper contributes novel taxonomy by introducing hybrid approach, which combines two leverage their respective advantages. Furthermore, review offers guidance for future research topic. Later discussion, three approaches estimating will broken down specific used, namely image processing techniques, machine learning models, deep or combination them. This also provides coherent discussion details these papers, well advantages drawbacks. To best our knowledge, first specifically discusses exclusively video data.

Language: Английский

Citations

0