Leveraging Computer Vision for Emergency Vehicle Detection-Implementation and Analysis

Kaushik, S. Sivaram; Raman, Abhishek; Rao, K. V. S. Rajeswara

doi:10.1109/icccnt49239.2020.9225331

Cited by 17 publications

(6 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared to the results reported in the state of the art with similar works, it is found that by means of the training parameters of the ResNet exposed in section 2, 94% of true positives are obtained compared to 74% of those reported in [24]. One reason for this is found in the database used in the training, for our case the images come from the detection from the point of view of the traffic light, where the ambulance is in front approaching it, while in [24] side views of the ambulance are also included, which in general helps the dispersion of the learning of the network, reducing its accuracy.…”

Section: Resultsmentioning

confidence: 63%

“…Among the applications of convolutional neural Int J Elec & Comp Eng ISSN: 2088-8708  networks (CNN), vehicle detection [21] stands out, mainly by means of faster region-based convolutional neural network (R-CNN) architectures [22], [23], based on regions (R), which allow locating the location of the vehicle in the image, facilitating, among others, the counting of vehicles. In [24], the detection of emergency vehicles by means of faster R-CNN networks and residual architectures based on regions (ResNet) is presented, showing the best performance in the latter case.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Ambulance detection for smart traffic light applications with fuzzy controller

Moreno

Baquero²,

Umaña³

2022

IJECE

View full text Add to dashboard Cite

<p>In the development of intelligent cities, the automation of vehicular mobility is one of the strong points of research, where intelligent traffic lights stand out. It is essential in this field to prioritize emergency vehicles that can help save lives, where every second counts in favor of the transfer of a patient or injured person. This paper presents an artificial intelligence algorithm based on two stages, one is the recognition of emergency vehicles through a ResNet-50 and the other is a fuzzy inference system for timing control of a traffic light, both lead to an intelligent traffic light. An application of traffic light vehicular flow control for automatic preemption when detecting emergency vehicles, specifically ambulances, is oriented. The training parameters of the network, which achieves 100% accuracy with confidence levels between 65% with vehicle occlusion and 99% in direct view, are presented. The traffic light cycles are able to extend the green time of the traffic light with almost 50% in favor of the road that must yield the priority, in relation to not using the fuzzy inference system.</p>

show abstract

Section: Resultsmentioning

confidence: 63%

Section: Introductionmentioning

confidence: 99%

Ambulance detection for smart traffic light applications with fuzzy controller

Moreno

Baquero²,

Umaña³

2022

IJECE

View full text Add to dashboard Cite

show abstract

“…Figure 2. Tabular representation of accuracy [3] The paper "A New Hybrid Architecture for Real-Time Detection of Emergency Vehicles" [4] presents a two-stage hybrid algorithm for emergency vehicle detection that combines machine learning and image processing. Operating at five frames per second using ResNet and the YOLO algorithm, the model crops the region of interest, applying an OCR module to detect "Ambulance" text, and can flip mirrored images for improved recognition.…”

Section: Related Workmentioning

confidence: 99%

A Comparative Study of Neural Networks for Emergency Vehicle Detection in Traffic

Palande,

Coutinho

2023

IJSREM

View full text Add to dashboard Cite

This study explores the application of Neural Networks for efficient emergency vehicle identification. Utilizing a dataset of emergency and non-emergency vehicle images, the model demonstrates high accuracy in classification. A noteworthy use case involves real-time integration with traffic systems, where the identification of an emergency vehicle triggers an automatic change of traffic lights to green, facilitating swift and unimpeded passage.

show abstract

“…The most common sensor used for highway monitoring is a vision sensor [ 1 ], such as a camera commonly found on highways. Although video image processing technology is a new traffic detection method that has been gradually developed in recent years, it is wireless, flexible in use, can detect multiple traffic parameters at once, and has a large detection range, and it has a very broad application prospect with the rise of high-definition cameras, deep learning, artificial intelligence, and other technologies [ 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 ]. However, the detection performance of vision sensors is significantly reduced under bad weather conditions such as low light, rain, snow, and fog.…”

Section: Introductionmentioning

confidence: 99%

Multi-Target Tracking Algorithm Combined with High-Precision Map

Cai

Zhu

et al. 2022

Sensors

View full text Add to dashboard Cite

On high-speed roads, there are certain blind areas within the radar coverage, especially when the blind zone is in curved road sections; because the radar does not have the measurement point information in multiple frames, it is easy to have a large deviation between the real trajectory and the filtered trajectory. In this paper, we propose a track prediction method combined with a high-precision map to solve the problem of scattered tracks when the targets are in the blind area. First, the lane centerline is fitted to the upstream and downstream lane edges obtained from the high-precision map in certain steps, and the off-north angle at each fitted point is obtained. Secondly, the normal trajectory is predicted according to the conventional method; for the extrapolated trajectory, the northerly angle of the lane centerline at the current position of the trajectory is obtained, the current coordinate system is converted from the north-east-up (ENU) coordinate system to the vehicle coordinate system, and the lateral velocity of the target is set to zero in the vehicle coordinate system to reduce the error caused by the lateral velocity drag of the target. Finally, the normal trajectory is updated and corrected, and the normal and extrapolated trajectories are managed and reported. The experimental results show that the accuracy and convergence effect of the proposed methods are much better than the traditional methods.

show abstract

Leveraging Computer Vision for Emergency Vehicle Detection-Implementation and Analysis

Cited by 17 publications

References 7 publications

Ambulance detection for smart traffic light applications with fuzzy controller

Ambulance detection for smart traffic light applications with fuzzy controller

A Comparative Study of Neural Networks for Emergency Vehicle Detection in Traffic

Multi-Target Tracking Algorithm Combined with High-Precision Map

Contact Info

Product

Resources

About