Urban Road Congestion Recognition Using Multi-Feature Fusion of Traffic Images

Cui, Hua; Li, Pannong; Wei, Zefa; Song, Xingguo; Guo, Lu

doi:10.23977/jaip.2016.11002

Cited by 1 publication

(1 citation statement)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Extracted various image features from traffic images such as sift corner, gray histogram variance, gray level cooccurrence matrix of energy, and contrast. The extracted features are used to train a neural network to classify traffic congestion [14]. Impedovo et al [15] analyzed several states of art object detectors, visual features and classification models used in traffic state estimations.…”

Section: Related Workmentioning

confidence: 99%

Flow Analysis of Vehicles on a Lane Using Deep Learning Techniques

Joshi,

Kulkarni

2023

JAIT

View full text Add to dashboard Cite

Increased volume of vehicles on the road everywhere in the world. It becomes difficult to regulate the traffic manually across all the traffic signals throughout the cities for the entire day. It is costly and tedious for manual traffic control at traffic signals. An Intelligent Transportation System (ITS) is proposed as a solution for traffic management and to address various challenges caused by increased vehicular density. Traffic analysis from real time camera images is the most adopted method for traffic state estimations which is a most important component in smart traffic management. The current traffic density estimators from videos estimate the volume of vehicles or percentage of area occupied in the lane. But for intelligent traffic management, more accurate traffic state estimation in terms of classification, density, speed, and flow rate for different categories of vehicles in different segments of the lane is needed. This work introduces a deep learning-based fine-grained vehicle flow analysis from traffic videos. A finegrained traffic density distribution for different categories of vehicles over the entire coverage area of the lane with flow information is referred to as a dynamic traffic state map. A continuous traffic state map integrating vehicle categorization, lane density estimation, vehicle flow estimation, and orientation flow analysis. Vehicle categorization based on high and low-level features is proposed instead of area-based thresholding. A novel deep learning-based vehicle density estimation integrating coherence-based region segmentation with convolutional neural network and density estimate from the segment is proposed. The solution is able to provide an estimate of the traffic at a fine-grained level.

show abstract

Section: Related Workmentioning

confidence: 99%