Vehicles Detection of Traffic Flow Video Using Deep Learning

Lou, Lu; Zhang, Qi; Sheng, Minglan; Zheng, Yu; Li, Xuan

doi:10.1109/ddcls.2019.8908873

Cited by 15 publications

(5 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The performed test on vehicle counting compared to several methods such as (Yang and Qu, 2018) which used background subtraction and Kalman filter to track detected vehicles in a video and achieved average counting accuracy of 92.2 %. (Lou et al, 2019) which used yolov3 with modified Kalman filter obtained average counting accuracy of 92.1 %. (Bhaskar and Yong, 2014) which used GMM and Blob detection and achieved average counting accuracy of 91%.…”

Section: Discussionmentioning

confidence: 99%

“…In addition to above modified CNN architecture (Al-Ariny et al, 2020)used improved Mask R-CNN for dealing the issues of congestion and occlusion. (Lou et al, 2019, Cepni et al, 2020, Sudha and Priyadarshini, 2020, Ligayo et al, 2021, Zuraimi and Zaman, 2021) used yolo-v3 as a novel object detection technique for vehicle detection. (Gupta et al, 2022)used tiny-yolov3 for military vehicle detection and classification in real-time environment, while (Song et al, 2019)used yolo-v3 to solve the issues of small object detection and multi-scale variation of the object in highway management, and (Rashmi and Shantala, 2020)used for analyzing vehicle density in urban roads and (Jin et al, 2021)to analyze urban traffic data in urban logistics.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Vehicle Detection, Counting, and Classification System based on Video using Deep learning Models

2024

ZJPAS

View full text Add to dashboard Cite

Traffic analysis is one of the crucial tasks of intelligent transport system that utilizes deep learning for range of purposes. Many tasks, such as vehicle recognition, vehicle counting, traffic violation monitoring, vehicle speed monitoring, vehicle density and so on, can be accomplished by using cameras installed in strategic locations along roads. In this paper powerful deep learning techniques such as (Yolov5, Mask R-CNN, SSD) and state-of-the-art object tracking algorithm known as DeepSORT was used to perform real time vehicle detection and counting in a video. A new highway vehicle detection dataset with overall of 32,265, instances of four vehicle classes named: bus, car, motorbike, truck was created in this paper and utilized for training vehicle detection and counting system. Result shows that average counting accuracy by using Yolov5 combined DeepSORT reaches to 95% while reaches to 91% by using Mask R-CNN combined DeepSORT and 84% by using SSD combined DeepSORT in hard environment. From the experimental work, counting accuracy by using Yolov5 outperforms other two deep learning techniques.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Vehicle Detection, Counting, and Classification System based on Video using Deep learning Models

2024

ZJPAS

View full text Add to dashboard Cite

show abstract

“…The system performed admirably under various lighting situations. The vehicle detecting system was proposed by [Lou et al (2019)]. YOLOv3 was utilized to detect moving cars, and the Kalman filter method was employed to follow a detected vehicle.…”

Section: Related Workmentioning

confidence: 99%

“…The vehicle is tracked after it is detected to count. Regarding vehicle tracking, earlier studies used the ORB algorithm [Song et al, (2019)], CSRT tracking model [Huy and Duc, (2020)], Deep SORT [Zuraimi and Zaman, (2021)], [Liu and Juang, (2021)], [Bui and Cho, (2020)], [Doan and Truong, (2020)], Hungarian algorithm [Lin and Jhang, (2022)], Kalman filter algorithm [Lou et al, (2019)]. For vehicle counting, the trajectory tracking method, TSI method [Yang et al, (2021)], and Virtual Detection Line [Bhuiyan et al, (2019)] were used.…”

Section: Vehicle Tracking and Countingmentioning

confidence: 99%

Traffic Monitoring System for Smart City Based on Traffic Density Estimation

Bidwe¹,

Kale²,

Bidwe³

2022

INDJCSE

View full text Add to dashboard Cite

Nowadays, traffic monitoring systems are at the frontline of smart city movement, and traffic density estimation is useful to a traffic monitoring system. The system of this work estimates traffic density using a five-layered CNN with a variety of input feature maps and filter sizes. There are 64, 64, 96, 96, and 96 feature maps for each pair of convolutions and max-pooling layers, and each pair's corresponding filter sizes are 5*5, 3*3, 5*5, 3*3, and 3*3. The proposed system divides the traffic into three categories: High, Medium, and Low traffic based on images that are taken from traffic videos that are recorded by traffic surveillance cameras. To test the system, we used the WSDT (Washington State Department of Traffic Transportation) Dataset of recorded video footage from the highway CCTVs in cities Seattle and Washington. The model is evaluated using parameters such as 0.99 precision, 0.99 recall, 0.99 f1-score, and model accuracy of 99.6 %. By examining the dataset, we have trained the model in such a manner to produce better results.

show abstract

“…YOLO V3 was also used in [53], reaching an 86% accuracy for vehicle detection. In [54], a system capable of detecting and tracking vehicles using YOLO V3 was proposed. In the tests, it reached an mAP of 92.11% in vehicle counting on congested roads at a speed of 2.55 FPS.…”

Section: Urban Traffic Solutions Using Different Versions Of the Yolomentioning

confidence: 99%

Lightweight PVIDNet: A Priority Vehicles Detection Network Model Based on Deep Learning for Intelligent Traffic Lights

Barbosa

Ayub

Rosa

et al. 2020

Sensors

View full text Add to dashboard Cite

Minimizing human intervention in engines, such as traffic lights, through automatic applications and sensors has been the focus of many studies. Thus, Deep Learning (DL) algorithms have been studied for traffic signs and vehicle identification in an urban traffic context. However, there is a lack of priority vehicle classification algorithms with high accuracy, fast processing, and a lightweight solution. For filling those gaps, a vehicle detection system is proposed, which is integrated with an intelligent traffic light. Thus, this work proposes (1) a novel vehicle detection model named Priority Vehicle Image Detection Network (PVIDNet), based on YOLOV3, (2) a lightweight design strategy for the PVIDNet model using an activation function to decrease the execution time of the proposed model, (3) a traffic control algorithm based on the Brazilian Traffic Code, and (4) a database containing Brazilian vehicle images. The effectiveness of the proposed solutions were evaluated using the Simulation of Urban MObility (SUMO) tool. Results show that PVIDNet reached an accuracy higher than 0.95, and the waiting time of priority vehicles was reduced by up to 50%, demonstrating the effectiveness of the proposed solution.

show abstract

Vehicles Detection of Traffic Flow Video Using Deep Learning

Cited by 15 publications

References 7 publications

Vehicle Detection, Counting, and Classification System based on Video using Deep learning Models

Vehicle Detection, Counting, and Classification System based on Video using Deep learning Models

Traffic Monitoring System for Smart City Based on Traffic Density Estimation

Lightweight PVIDNet: A Priority Vehicles Detection Network Model Based on Deep Learning for Intelligent Traffic Lights

Contact Info

Product

Resources

About