Vehicle Classification and Counting from Surveillance Camera Using Computer Vision

There is a drastic increase experienced in the production of vehicles in recent years across the globe. In this scenario, vehicle classification system plays a vital part in designing Intelligent Transportation Systems (ITS) for automatic highway toll collection, autonomous driving, and traffic management. Recently, computer vision and pattern recognition models are useful in designing effective vehicle classification systems. But these models are trained using a small number of hand-engineered features derived from small datasets. So, such models cannot be applied for real-time road traffic conditions. Recent developments in Deep Learning (DL)-enabled vehicle classification models are highly helpful in resolving the issues that exist in traditional models. In this background, the current study develops a Lightning Search Algorithm with Deep Transfer Learning-based Vehicle Classification Model for ITS, named LSADTL-VCITS model. The key objective of the presented LSADTL-VCITS model is to automatically detect and classify the types of vehicles. To accomplish this, the presented LSADTL-VCITS model initially employs You Only Look Once (YOLO)-v5 object detector with Capsule Network (CapsNet) as baseline model. In addition, the proposed LSADTL-VCITS model applies LSA with Multilayer Perceptron (MLP) for detection and classification of the vehicles. The performance of the proposed LSADTL-VCITS model was experimentally validated using benchmark dataset and the outcomes were examined under several measures. The experimental outcomes established the superiority of the proposed LSADTL-VCITS model compared to existing approaches.

The rapid advancement in the field of deep learning and high performance computing has highly augmented the scope of video-based vehicle counting system. In this paper, the authors deploy several state-of-the-art object detection and tracking algorithms to detect and track different classes of vehicles in their regions of interest (ROI). The goal of correctly detecting and tracking vehicles’ in their ROI is to obtain an accurate vehicle count. Multiple combinations of object detection models coupled with different tracking systems are applied to access the best vehicle counting framework. The models’ addresses challenges associated to different weather conditions, occlusion and low-light settings and efficiently extracts vehicle information and trajectories through its computationally rich training and feedback cycles. The automatic vehicle counts resulting from all the model combinations are validated and compared against the manually counted ground truths of over 9 h’ traffic video data obtained from the Louisiana Department of Transportation and Development. Experimental results demonstrate that the combination of CenterNet and Deep SORT, and YOLOv4 and Deep SORT produced the best overall counting percentage for all vehicles.

The following paper depicts an architectural layout and operational insights towards an optoelectronic system dedicated for vehicle counting and classification, designed to monitor occupancy of rest areas. The system is based upon an image cue acquisition by video cameras and advanced signal processing mechanism. In order to satisfy the appropriate realization of measurement procedure (counting and classification), a specific measurement configuration scheme alongside optimized video stream capturing parameters were set up. Concerning signal processing methods, the most effective and appropriate for the purpose of vehicle counting and classification were selected. Moreover, the article presents the outcomes of the designed measurement campaign conducted on the S7 expressway within the agenda of the National Center for Research and Development RID project. The results obtained with use of the designed methodology were compared to the outcomes of the manual method.

Vehicle Tracking and Counting through Image Processing is an efficient system for real-time vehicle detection, classification, and counting using the powerful combination of YOLO and OpenCV. The System uses the application of machine learning in the domain of image processing for vehicle tracking and counting. OpenCV in this system is used to process the video feed or images and provide real-time visualization of the vehicles for detection. The system uses advanced computer vision techniques such as contour detection, image segmentation, and feature extraction to accurately locate and recognize individual vehicles in a scene. The detected vehicles are classified into various categories, including cars, trucks, motorcycles, and bicycles (Two & four wheelers) using the trained YOLO model. In addition to detection and classification the count of vehicles is also noted down. The primary objective of this system is to track the vehicles, distinguish and count them. This system can be used in – Traffic Signal Management and Parking Lot Monitoring. Keywords: - YOLO, OpenCV, Python, CVAT

Accurate vehicle count and classification reporting is a principal application of intelligent transportation system. United States Departments of Transportation (DOTs) use this information to design safe, suitable roadways. A wide variety of classification systems are currently employed. Regardless of type, an optimal system must be cost effective, require minimal maintenance, cause limited pavement damage during installation, and provide accurate information. Currently, the most widely used systems combine inductive loops and piezoelectric sensors. Most cause severe pavement damage during installation and fail to provide accurate classification for a variety of classes-class 1 motorcycles in particular. This paper introduces a vehicle classification system that employs a single element piezoelectric sensor placed diagonally across a traffic lane. The system is able to accurately classify 13 Federal Highway Administration vehicle classes, including class 1 motorcycles. The novel classification method utilizes the ratio of vehicle track width-to-length (w/l) as well as a traditional axle-spacing scheme to classify vehicles. The single-element system is cost effective and reduces pavement damage primarily because it utilizes only a single sensor. Two highway test deployments found the system capable of achieving classification accuracy of up to 98.8% when employing the (w/l) method and 84.4% when employing the axle-spacing method.

This paper explores the feasibility of using a lowcost embedded ARM-based system for real-time vehicle detection, classification and counting through image processing algorithms with the aim of knowing information about vehicular traffic in different roads and highways to improve the management of mobility and the functioning of cities. This paper proposes the implementation of a low cost system to identify and classify vehicles using an Embedded ARM based platform (ODROID XU-4) with Ubuntu operating system. The algorithms used are based on the Open-source library (Intel OpenCV) and implemented in Python programming language. The experimentation carried out proved that the efficiency of the algorithm implemented was 95.35%, but it can be improved by increasing the training sample.

The paper presents the existing situation of automatic traffic recording systems located on the national road and motorway network in Romania, highlighting their technical and operational limitations, as well as the need to modernize and expand the traffic counter network. In addition, the implementation of an innovative vehicle counting and classification system based on the Raspberry Pi platform – VEKPI-count is developed and tested within a pilot project. The operating principles, technical and economic advantages of this system are presented, alongside comparative results obtained against conventional equipment currently in use. The results of the study indicate that the use of the VEKPI-count system leads to an improvement in the quality of traffic data and a reduction in classification errors, while providing a cost-effective solution for expanding the monitoring network. The large-scale implementation of such systems can significantly contribute to the analysis of the correlation of road accidents with road characteristics and to increasing traffic safety.

Artificial intelligent systems for vehicle classification: A survey

Using video in traffic monitoring is one of the most active research domains in the computer vision community. TrafficMonitor, a system that employs a hybrid approach for automatic vehicle tracking and classification on highways using a simple stationary calibrated camera, is presented. The proposed system consists of three modules: vehicle detection, vehicle tracking, and vehicle classification. Moving vehicles are detected by an enhanced Gaussian mixture model background estimation algorithm. The design includes a technique to resolve the occlusion problem by using a combination of two-dimensional proximity tracking algorithm and the Kanade–Lucas–Tomasi feature tracking algorithm. The last module classifies the shapes identified into five vehicle categories: motorcycle, car, van, bus, and truck by using three-dimensional templates and an algorithm based on histogram of oriented gradients and the support vector machine classifier. Several experiments have been performed using both real and simulated traffic in order to validate the system. The experiments were conducted on GRAM-RTM dataset and a proper real video dataset which is made publicly available as part of this work.

Vehicle type classification based on images is widely applied in traffic surveillance and monitoring system. It has been used in billing at toll collection stations and preventing heavy trucks from entering city viaducts. A hierarchical vehicle detection and classification system is proposed in this paper. A cascade ensemble classifier, accepting Multiple Layer Perceptron (MLP) and K-Nearest Neighbor (K-NN) as base classifiers, is proposed for the vehicle type classification. The experiments are conducted, and the hierarchical classifier with two layers offers a reliability of 97.8% with a rejection rate of 3.0%, which show the effectiveness of our proposed hierarchical vehicle detection and classification system.

This survey paper presents a comprehensive overview of AI-based traffic detection and autonomous signal operation systems, focusing on the integration of advanced technologies such as cameras, Raspberry Pi, and image processing algorithms, particularly YOLO (You Only Look Once). With the rapid urbanization and the increasing number of vehicles, efficient traffic management has become crucial for improving road safety and reducing congestion. Our survey examines various methodologies employed in traffic detection, including real-time monitoring and classification of vehicles, pedestrians, and cyclists. We analyze the effectiveness of using Raspberry Pi as a low-cost computing platform in conjunction with high-resolution cameras for image acquisition and processing. The paper highlights the performance of YOLO in detecting traffic objects in diverse environments, showcasing its speed and accuracy in real-time applications. Furthermore, we discuss the challenges faced in deploying these systems in real-world scenarios, including environmental variability, data privacy concerns, and the need for robust infrastructure. Finally, we propose future directions for research, emphasizing the importance of integrating machine learning with IOT solutions to create smarter, more responsive traffic management systems. This survey aims to provide insights for researchers and practitioners in developing sustainable traffic solutions leveraging AI technologies.

In today’s world, a commonly used computer technology - object detection focuses on finding items or examples of a specific class such as humans, vehicles, and animals in digital photos. Most computer vision systems require the capacity to detect objects. Object detection is performed often in computer vision tasks such as face detection, face identification, and video object co-segmentation, as well as in tracking people, surveillance cameras, and driverless cars. It is the fundamental notion for tracking and identifying things, and it has an impact on object recognition efficiency and accuracy. The major contribution of this paper is i) Comparison of recent research works in the domain of vehicle detection and ii) Classification of vehicles into 2 specified classes – car and bus are performed on a dataset of 10,000 images. The YOLOv4 model has been deployed for detecting cars and buses.

In this paper we describe holonic organization of a multi agent system for automatic vehicle classification in a road toll system. Classification of vehicles in road toll systems is based on physical vehicle features and in this paper we focus on axle counting as the first discriminant feature for class determination. Our system relies on two main sensors—video camera and depth sensor. Video image and depth image processing is performed in several holons. The results from individual holons are fused into the final decision on a number of axles of a passing vehicle. We show that fusion of results from individual holons gives more precise results than individual holons. Holonic organization of the system aids scalability and simplifies inclusion of new sensors and new algorithms.

SummaryThis paper presents novel vehicle classification technology by utilizing a single‐element piezoelectric sensor placed diagonally on a traffic lane to accurately identify vehicles. Novelty of this technique originates from using diagonally placed piezoelectric strip sensor and machine learning technology to provide a highly accurate and cost‐effective alternative to current vehicle classification systems. Diagonal placements of the piezoelectric strip sensor ensure detection of passing vehicle tires by facilitating vehicle classification process. Presented technology is capable of accurately classifying vehicles into a relatively large number of classifications, including motorcycle, which has proven to be a challenging category in present‐day commercial vehicle classifiers. Vehicle classification is a vital intelligent transportation systems application. Accurate data reporting aids suitable roadway design for safety and capacity and can also support other purposes, such as reporting highway congestion to the general public or providing area denseness data to interested businesses.To make a classification decision, a vehicle's signal is acquired from diagonal piezoelectric strip sensor, processed, and then applied to a machine learning algorithm. A speed estimation technique using the same single‐element piezoelectric sensor was also developed, tested, and compared with an embedded vehicle classifier currently used by the Oklahoma Department of Transportation. Testing on several highway sites indicated up to 97% classification accuracy. This paper presents a complete description of the developed system, including sensor installation, data acquisition and processing, and classification algorithm. Overall, the system offers a high‐performance cost‐effective solution for vehicle classification that minimizes roadwork typically required for loop and sensor installations of current systems. Copyright © 2016 John Wiley & Sons, Ltd.

This paper provides a review of a 3-year research program on the feasibility of using airborne LiDAR (Light Detection and Ranging) and imagery collected simultaneously over transportation corridors for estimation of traffic flow parameters such as: (1) vehicle counts, (2) vehicle classification, (3) velocity per vehicle category, and (4) intersection movement patterns. This work is conducted by The National Consortium for Remote Sensing in Transportation-Flows (NCRST-F), led by The Ohio State University, supported by the U.S. Department of Transportation and the National Aeronautics and Space Administration (NASA). The major focus is on improving the efficiency of transportation systems by integration of remotely sensed data with traditional ground data to monitor and manage traffic flows.