Download fulltext HTML

https://doi.org/10.18502/espoch.v2i5.11734

statistics

  • 451 Abstract Views
  • 155 PDF Views
  • 0 HTML Views

authors

  • P. Arroba-Villacis

    P. Arroba-Villacis

    Instituto Superior Pedagógico Martha Bucaram de Roldós Bilingüe – Nueva Loja, Ecuador
  • W. Maiza Pérez

    W. Maiza Pérez

    Instituto Superior Pedagógico Martha Bucaram de Roldós Bilingüe – Nueva Loja, Ecuador
  • C. Carrión-Paladines

    C. Carrión-Paladines

    Instituto Superior Pedagógico Martha Bucaram de Roldós Bilingüe – Nueva Loja, Ecuador
  • F. Revelo-Aguilar

    F. Revelo-Aguilar

    Universidad Técnica de Ambato, Ambato- Ecuador

Published Date

  • Sep 1, 2022

Vehicle and Pedestrian Detection in Traffic Videos Using Convolutional Neural Networks

ESPOCH Congresses: The Ecuadorian Journal of S.T.E.A.M. / Volume 2, Issue 5 / Pages 1301–1312

Abstract

One of the major applications of computer vision is the analysis of the traffic scene on the road, and how pedestrian traffic affects traffic in general. Road sizes and traffic signals must constantly adapt. Counting and classifying vehicles and pedestrians at an intersection is an exhausting task, and despite the use of traffic control systems, human interaction is very necessary to perform such a task. The object of study of Deep Learning is to try to solve problems that require artificial intelligence. Artificial intelligence has been working in this field for years, with different approaches and algorithms. It has achieved an important emergence in the recognition of patterns in images and videos using these techniques, to the point of surpassing human capacity in some problems. An important factor in this development is the ability to process large volumes of information in applications, which has resulted in the devices used for this purpose, such as GPU’s and multi-core CPU’s, requiring a large amount of power to operate. For the development of the application of vehicle and pedestrian detection in traffic videos, YOLO V3 was used, which is a neural network model of the latest generation of real-time objects.


Keywords: yoloV3, Deep Learning, Convolucional Network.


Resumen


Una de las mayores aplicaciones de la visión por computadora es el análisis de la escena de tráfico en la carretera, y cómo el tráfico de peatones afecta al tráfico en general. Los tamaños de las carreteras y las señales de tráfico deben adaptarse constantemente. Contar y clasificar vehículos y peatones en una intersección es una tarea agotadora y, a pesar del uso de sistemas de control de tráfico, la interacción humana es muy necesaria para realizar dicha tarea. El objeto de estudio de Deep Learning, es intentar resolver problemas que requieren inteligencia artificial. La inteligencia artificial ha trabajado en este campo durante años, con diferentes enfoques y algoritmos. Ha logrado un surgimiento importante en el reconocimiento de patrones en imágenes y videos usando estas técnicas, hasta el punto de superar la capacidad humana en algunos problemas. Un importante factor de este desarrollo es la capacidad de procesar grandes volúmenes de información en aplicaciones, lo que ha dado como resultado que los dispositivos utilizados para este propósito, como GPU’s y CPU’s multinúcleo, requieran una gran cantidad de energía para operar. Para el desarrollo de la aplicación de Detección de vehículos y peatones en videos de tráfico, fue utilizado YOLO V3, que es un modelo de red neuronal de la última generación de objetos en tiempo real.


Palabras Clave: yoloV3, Aprendizaje profundo, Red convolucional

References
[1] Poppe R. A survey on vision-based human action recognition. Image and vision Computing. 2010 June; 28(6).

[2] Andres Felip JIM. Detección de flujo vehicular basado en visión artificial. Scientia et Technica. 2007; 3(35).

[3] Rusk N. Deep learning. Nature Methods. 2016; 13(1).

[4] Bewley A,GZ,OL,RF,&UB. Simple online and realtime tracking. In IEEE international conference on image processing (ICIP); 2016. p. 3464-3468.

[5] Human fall detection algorithm based on YOLOv3. In 2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC). In IEEE 5th International Conference on Image, Vision and Computing (ICIVC) ; 2020. p. 50-54.

[6] He W,HZ,WZ,LC,&GB(. TF-YOLO: An improved incremental network for real-time object detection. Applied Sciences. 2019; 9(16).

[7] Zhang F,LC,&YF. Vehicle detection in urban traffic surveillance images based on convolutional neural networks with feature concatenation. Sensors. 2019; 19(3).

[8] Wojke N,&BA. Deep cosine metric learning for person reidenti. In IEEE winter conference on applications of computer vision; 2018. p. 748-756.

[9] Gong J,ZJ,LF,&ZH. Vehicle detection in thermal images with an improved yolov3- tiny. In IEEE international conference on power, intelligent computing and systems (ICPICS); 2020. p. 253-256.

[10] Benjdira B,KT,KA,AA,&OK. Car detection using unmanned aerial vehicles: Comparison between faster r-cnn and yolov3. In 1st International Conference on Unmanned Vehicle Systems-Oman (UVS); 2019. p. 1-6.

[11] Huang YQ,ZJC,SSD,YCF,&LJ. Optimized YOLOv3 algorithm and its application in traffic flow detections. Applied Sciences. 2020; 10(9).

[12] Zhang H,QL,LJ,GY,ZY,ZJ,&XZ. Real-time detection method for small traffic signs based on Yolov3. IEEE Access. 2020; 8.

[13] Zhang FK,YF,&LC. Fast vehicle detection method based on improved YOLOv3. Computer Engineering and Applications. 2020; 55(2).

[14] Hassan, N. I., Tahir, N. M., Zaman, F. H. K., & Hashim, H. In 10th IEEE International Conference on Control System, Computing and Engineering (ICCSCE); 2020. p. 131- 136.

[15] Ouyang L,&WH. Vehicle target detection in complex scenes based on YOLOv3 algorithm. In IOP Conference Series: Materials Science and Engineering; 2019. p. 052018.

[16] Wang X,WS,CJ,&WY. Data-driven based tiny-YOLOv3 method for front vehicle detection inducing SPP-net. IEEE Access. 2020; 8.

[17] Du L,CW,FS,KH,LC,&PZ. Real-time detection of vehicle and traffic light for intelligent and connected vehicles based on YOLOv3 network. In 5th International Conference on Transportation Information and Safety (ICTIS) ; 2019. p. 388-392.

[18] Zadobrischi E,&NM. Pedestrian detection based on TensorFlow YOLOv3 embedded in a portable system adaptable to vehicles. In International Conference on Development and Application Systems (DAS); 2020. p. 21-26.

[19] Zhao S,&YF. Vehicle detection based on improved yolov3 algorithm. In International Conference on Intelligent Transportation, Big Data & Smart City (ICITBS) ; 2020. p. 76-79.

[20] Zhou L,LJ,&CL. Vehicle detection based on remote sensing image of Yolov3. In IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference (ITNEC); 2020. p. 468-472.

[21] Hassan NI,TNM,ZFHK,&HH. People detection system using YOLOv3 algorithm. In IEEE International Conference on Control System, Computing and Engineering (ICCSCE); 2020. p. 131-136.

[22] Zhao H,ZY,ZL,PY,HX,PH,&C. Mixed YOLOv3-LITE: a lightweight real-time object detection method. Sensors. 2020; 20(7).

[23] Pérez RM,AJS,&PAM. Introducción al Aprendizaje Automático con YOLO. 2019; 3(6).

[24] Wojke N,BA,&PD. Simple online and realtime tracking with a deep association metric. In IEEE International Conference on Image Processing (ICIP); 2017. p. 3645-3649.

[25] Zhang X,&ZX. An efficient and scene-adaptive algorithm for vehicle detection in aerial images using an improved YOLOv3 framework. ISPRS International Journal of Geo-Information. 2019; 8(11)