TX_1~ABS:AT/TX_2:ABS~AT 6 UHD Journal of Science and Technology | Jan 2023 | Vol 7 | Issue 2 1. INTRODUCTION Number of cars has significantly increased nowadays, [1], [2] consequently, traffic congestion problem has been arise around the world [3]. Subsequently, vehicle clashing and crashing and dramatic increase of CO2 emission per year [4] are threatening sustainable mobility of future [5]. Further more, traffic control needs man power to be controlled [6]. The traffic control devices are time dependent and designed to flow the traffic in all directions. On top of that, sometimes during turning the lights from green to red causes traffic deadlock in a direction without having a noticeable flow in the other direction [7]. Congestions caused by traffic signals could negatively impact on economy in terms of transportation due to fuel [8], time expenditure [9], and air pollution [10]. Moreover, injuring even sometimes death caused by accidents happened in deadlock traffics [8], on the other hand, reducing congestion may have economic, environmental, and social benefits. In general, to make the optimization problem manageable, several assumptions have to be made. The main problem that arises is that these assumptions deviate and sometimes do so significantly from the real world. Meanwhile, many factors have effects on drivers in real world traffics such as on driver’s preference interactions with vulnerable road users (e.g., pedestrians, cyclists, etc.), weather and road conditions [11]. On the other hand, computer vision has an important role in managing and controlling traffic signals with great success [6], [12]. The best way to control traffic flow in big and busy cities is to utilize intelligent traffic signal [6], the system has ability to approximately evaluate density estimation, A Review of Computer Vision–Based Traffic Controlling and Monitoring Kamaran H. Manguri1,2, Aree A. Mohammed3 1Department of Technical Information Systems Engineering, Erbil Technical Engineering College, Erbil Polytechnic University, Erbil, Iraq, 2Department of Computer Science, College of Basic Education, University of Raparin, Ranya 46012, Iraq, 3Department of Computer Science, College of Science, University of Sulaimani, Sulaymaniyah, Iraq A B S T R A C T Due to the rapid increase of the population in the world, traffic signal controlling and monitoring has become an important issue to be solved with regard to the direct relation between the number of populations and the cars’ usage. In this regard, an intelligent traffic signaling with a rapid urbanization is required to prevent the traffic congestions, cost reduction, minimization in travel time, and CO2 emissions to atmosphere. This paper provides a comprehensive review of computer vision techniques for autonomic traffic control and monitoring. Moreover, recent published articles in four related topics including density estimation investigation, traffic sign detection and recognition, accident detection, and emergency vehicle detection are investigated. The conducted survey shows that there is no fair comparison and performance evaluation due to the large number of involved parameters in the abovementioned four topics which can control the traffic signal controlling system such as (computation time, dataset availability, and an accuracy). Index Terms: Traffic signaling system, Intelligent traffic, Computer vision, Traffic congestion, Traffic monitoring, Review. Corresponding author’s e-mail:  kamaran@uor.edu.krd Received: 20-12-2022 Accepted: 26-04-2023 Published: 10-08-2023 Access this article online DOI: 10.21928/uhdjst.v7n2y2023.pp6-15 E-ISSN: 2521-4217 P-ISSN: 2521-4209 Copyright © 2023 Kamaran H. Manguri and Aree A. Mohammed. This is an open access article distributed under No Derivatives License 4.0 (CC BY-NC-ND 4.0) RE V I E W A R T I C L E UHD JOURNAL OF SCIENCE AND TECHNOLOGY Manguri and Mohammed: Traffic Controlling and Monitoring: A Review UHD Journal of Science and Technology | Jan 2023 | Vol 7 | Issue 2 7 traffic signals detection and recognition, emergency and police car detection, and accident detection. Even though a better infrastructure can improve the traffic flow [13]. Usually in quiet intersections, the traffic is controlled by human or system controls [6]. In most congestions, cameras have been put for purposes other than traffic control, such as security, vehicle detection, and arrangement [14]. These cameras can be utilized for the reason of analyzing traffic scenes simply by employing specific hardware. The main advantage is that there is no need for replacing the CCTVs. The main objective of this survey is to fill the research gap that exists in the field of traffic signal controlling and monitoring. The importance of this survey is to propose some technique based on computer vision for reducing the road congestion and keeping the environment green and public health. In this study, different approaches based on computer vision for traffic signaling controls are reviewed. For this purpose, the literatures over the period January 2015–January 2022 are surveyed. The structure of this review is as follows: Section I provides an introduction to the traffic and its problems. Background of traffic management addressed in Section II. In Section III, a literature review is provided for the existing solution of the intelligent traffic signaling. Section IV provides a discussion of review of the existing solutions. Finally, conclusion remarks are presented in last section. 2. REVIEW STRATEGY This review is aimed in analyzing the recent literature for the vision-based methods for traffic controlling and managing, which have been published from January 2015 to January 2022 in terms of journal papers and conference proceedings. The reviewed papers were chosen after an extensive manual search of databases including IEEE Xplore, Springer, Elsevier, and Google Scholar. Keywords used to explore the databases are shown in Table 1. In addition, vision or image processing keywords are selected as the main keywords in the title of papers. Moreover, one of the sub search keywords has been used with main keyword to find the studies in the above mentioned period. 3. URBAN TRAFFIC LIGHT SYSTEM Usually, each traffic light contains three color lights precisely, green, yellow, and red lights. They are put in the four parallel and perpendicular directions [15]. Fig. 1 shows a common intersection that formed by two perpendicular and parallel lanes. Globally, the meaning of the lights for the drivers is as follows, green light means that the current lane has right to move forward meanwhile all other three directions are red which means they are not allowed to flow [11]. Besides, models of controlling traffic signaling and monitoring using computer vision required CCTV camera to acquisition images from the live traffic intersection. The simulation of traffic controlling in the cross road is shown in Fig. 2 [16]. 4. LITERATURE REVIEW To improve traffic signaling control and monitoring, scientists and researchers proposed many methods based on machine vision. Computer vision-based architecture of traffic signaling controlling and monitoring includes image acquisition, preprocessing and applying advance computer vision techniques density estimation, traffic sign detection and recognition, accident detection, and emergency vehicle detection. In this review, papers are randomly selected according to proposed methods in the recent years (between January 2015 and January 2022) for controlling and monitoring traffic signals. 4.1. Density Estimation Density estimation is a key aspect for automatic traffic signaling control and reducing congestion in the intersection areas. Different approaches by reviewers to estimate traffic density are detailed below: Table 1: Search parameters of the literature review Date range Database Main Keywords (OR) AND/Sub Search Key-words January 2015–January 2022 IEEE Xplore Springer Elsevier Google Scholar Vision Image Processing Machine Learning Deep Learning Traffic Controlling Traffic Density Traffic Congestion Crowd Detection Accident Accident Detection Accident Identification Emergency Vehicle Traffic Sign Manguri and Mohammed: Traffic Controlling and Monitoring: A Review 8 UHD Journal of Science and Technology | Jan 2023 | Vol 7 | Issue 2 Garg et al. [17] presented the approach for estimating traffic density based on vision which forms the fundamental building block of traffic monitoring systems. Due to the low accuracy of vehicle counting and tracking of existing techniques, the sensitivity to light changes, occlusions, congestions, etc. are made. Moreover, the authors addressed another problem of existing holistic-based methods by difficulty of implementation in real-time because the high computational complexity is required. To handle this issue, density is calculated using block processing approach for busy road segments. The proposed method involves two steps including marking of region of interest (ROI), generating block of interest, and background construction in the first step. Recurring process has been applied in the second step which involves background update, occupied block detection, shadow block elimination, and traffic density estimation. Finally, the proposed methods are evaluated and tested using the TrafficDB dataset. In Biswas et al. [1] density estimated based counting cars, Background Subtraction (BS) method and OverFeat framework are implemented. The accuracy of the proposed system is evaluated by manual counting of cars. Furthermore, the comparative study was conducted before and after outperforming OverFeat framework. Average accuracy reached 96.55% after applied OverFeat Framework from 67.69% average accuracy for placemeter and 63.14% average accuracy for BS, respectively. Furthermore, this study confirmed that the OverFeat Framework has another application area. The advantages and shortcomings of the BS and six individual obtained traffic videos have used for analyzing OverFeat Framework with regarding different perspectives such as camera angles, weather conditions, and daily time. Biswas et al. [3] implemented single shot detection (SSD) and MobileNet-SSD for estimating traffic density. For this purpose, 59 individual traffic cameras used for analyzing the SSD and MobileNet-SSD framework advantages and shortcomings. Moreover, two algorithms are compared with manually estimated density. The SSD framework demonstrates significant potential in the field of traffic density estimation. According to their experiment, the significant accuracy of detection achieved, numerically speaking the precisions were 92.97% and 79.30% for SSD and MobileNet-SSD, respectively. Bui et al. [18] developed a method for analyzing traffic flow, advanced computer vision technologies have been used to extract traffic information. For finding traffic density estimation in intersections data acquired from video surveillance. Moreover, YOLO and DeepSORT techniques turned for the detection, tracking, and counting of vehicles have enveloped to estimate the road traffic density. To evaluated the proposed method, data collected in a real-world traffic through CCTV during 1 day. A new technique for estimating traffic density utilizing a macroscopic approach has been developed by Kurniawan et al. [19]. The proposed method contains two parts including background construction and a traffic density estimation algorithm. The background construction obtained from detected non-moved vehicles in the front or behind vehicles. Moreover, background of the image founded using the edge detection technique. Density estimated by founding the ration between the number of ROI containing object and the total number of ROI. Eamthanakul et al. [20] proposed a method-based image processing techniques for congestion detection. The Fig. 1. Four road lanes intersection. Fig. 2. Vision-based crossroad model. Manguri and Mohammed: Traffic Controlling and Monitoring: A Review UHD Journal of Science and Technology | Jan 2023 | Vol 7 | Issue 2 9 technique contains three parts: (1) Image background substation used for separating vehicles from the background, (2) Morphological techniques applied for removing the image noises, and (3) traffic density calculated from the obtained image from CCTV. Finally, the results of the process are sent to transport plan database. 4.2. Traffic Sign Detection and Recognition Traffic sign recognition plays a key role in driver assistance systems and intelligent autonomous vehicles. Furthermore, it can be helpful for automatic traffic signals which leads to prevent pass across the intersections in the case of read signals. Novel approaches proposed in Berkaya et al. [21] for traffic sign detection and recognition. A new method developed to detect traffic sign under the name of circle detection algorithm. In addition, RGB-based color thresholding technique was proposed by Berkaya et al. [21]. Moreover, three algorithms have been used to recognize traffic signs including histogram of oriented gradients (HOG), local binary patterns and Gabor features are employed within a support vector machine (SVM) classification framework. The performance of the proposed methods for both detection and recognition evaluated on German Traffic Sign Detection Benchmark (GTSDB) dataset. Based on the obtained results from experiments, the proposed system better than the reported literatures and can be used in a real-time operation. Yang et al. [22] presented a method for traffic sign detection and recognition, the method includes three steps. Thresholding of HSI color space components used to segment image in the first step. Applying the blobs extracted to the first step for detecting traffic signs in the second step. The contribution of their method in the first step, machine learning algorithms not used classify shapes instead of this invariant geometric moments have been used. Second, inspired by the existing features, new method has been proposed for the recognition. The HOG features have been extended to the HSI color space and combined with the local self-similarity (LSS) features to get the descriptor. As a classifier, random forest and SVM classifiers have been tested together with the new descriptor. GTSDB and the Swedish Traffic Signs (STS) data sets have been used to test the proposed system. Finally, the results of the presented technique compared with existing techniques. Salti et al. [23] combined solid image analysis and pattern recognition techniques for detecting traffic sign in mobile mapping data. The system designed base on interest regions extracting which makes a significant with other existing systems that sliding window detection have been used. Furthermore, with having challenging conditions such as varying illumination, partial occlusions, and large scale variations, the proposed system good performance demonstrated. Three variant category traffic signs aimed to detect including mandatory, prohibitory and danger traffic signs, according to the experimental setup of the recent GTSDB competition. With having a very good performance of the proposed method in the online competition, the proposed method challenging dataset mobile mapping of Italian signs the pipeline has been evaluated and showed its successfully be deployed in real-world mobile mapping data. In Du et al. [24] designed the robust and fast performance classifier-based detector. They addressed two algorithms for detection and classification. First, aggregate channel features based on three types of features, which including the color feature, the gradient magnitude, and gradient histograms proposed. Second, boosted trees classifier multiscale and multiphase detector have been proposed based on Real AdaBoost algorithm. The obtained results from experiments of this study show high average-recall and speed which is evaluated on Daimler, LISA, and LaRA datasets. Real-time traffic signs’ detection and recognition are necessary for smart vehicles to make them more intelligent. To deal with this this issue. Shao et al. [25] are proposed a new approach that includes two steps; in the first one acquitted images from the road scene converted to grayscale images. Then simplified Gabor wavelets (SGW) filter has been applied to the optimized parameters of grayscale images. Furthermore, traffic sings bounded by edge detection which helps preparing the obtained result to the next process. In the second, the ROI extracted using the maximally stable extremal regions algorithm and the superclass of traffic signs are classified by SVM. To classify their subclasses, the traffic signs convolution neural networks (CNN) with input by simplified Gabor feature maps, where the parameters were the same as the detection stage is used. Finally, the proposed method tested on GTSDB and CTSD datasets and the results obtained from the experiments show that the method is fast and accurate by 6.3 frames per second and 99.43%, respectively. Berkaya et al. [21] presented new ideas to provide colorful graphics to improve traffic in terms of object recognition and problem detection. Two digital image processing methods, namely, Circle Detection Algorithm and RGB which based on the simplest image segmenting method have been improved to develop the ability of traffic sign Manguri and Mohammed: Traffic Controlling and Monitoring: A Review 10 UHD Journal of Science and Technology | Jan 2023 | Vol 7 | Issue 2 detection. The classification framework, namely, SVM has been formed through assembling three main attributes including Gabor features HOG, and local binary patterns in the smart system. The presented technique is validated by German Traffic Sign Detection and Recognition Benchmark datasets, correspondingly. According to the practical results, their technique is by far more efficient than the quoted approaches in this paper; the results are also aligned with the real time operation. A new approach for detecting and recognizing traffic signs proposed in Ellahyani et al. [26] which includes three main steps. Thresholding of HIS has been used to segment the image based on components of color spaces in the first step. It followed by applying blobs by the result of extracted from the former step. Then, the traffic signs recognition performed for the detected signs in the last step. Moreover, in their study, two different approaches used to classify signs. Instead of machine learning algorithms, invariant geometric moments used to classify shapes in the first step. Second, inspired by the existing features, new ones have been proposed for the recognition. HSI color space taken from the HOG features and combined with the LSS features to get the descriptor while used in the proposed algorithm. Then, last test has been done based machine learning algorithms which are random forest and SVM classifier. Finally, the performance of proposed method evaluated and tested on German Traffic Sign Recognition Benchmark (GTSRB), GTSDB, and STS datasets. Convolutional Neural Networks (CNN) machine learning algorithm is applicable for object recognition by having power full recognition rate and less time required for execution. In Shustanov and Yakimov [27] implemented traffic sign recognition using CNN. Furthermore, several architectures of CNN compared together. Meanwhile, Tensor Flow library is used for training and massively parallel architecture for multithreaded programming CUDA. The entire procedure for traffic sign detection and recognition is executed in real time on a mobile GPU. Finally, their method efficiency evaluated on GTSRB dataset and it is obtained very good result by 99.94% for classification images. 4.3. Accident Detection A main aspect of traffic monitoring is the identification and tracking of vehicles. Monitoring vehicles helps to report and detect in the situation of the traffic junctions. One of the main aspects of traffic monitoring is the identification and tracking of vehicles. In this section, accident prediction and detection approaches are faced. Tian et al. [28] developed a Cooperative Vehicle Infrastructure Systems (CVIS) and proposed machine based-vision that can be used to detect car accident automatically. The study includes two phases; CAD-CVIS database has been created to improve the accuracy of accident detection in the first phase. CAD-CVIS dataset with regarding different traffic situations consists of various types of accidents, weather conditions and accident location. In the second phase, to detect accident deep neural network model YOLO-CA based on CAD-CVIS and deep learning algorithms developed. Moreover, to improve the performance of the model for detection small objects Multiscale Feature Fusion and loss function with dynamic weights utilized. The results showed the proposed method faster than the previous methods, it can detect car accident in milliseconds with a very good average precision by 90.02%. Finally, the proposed methods compared with existing methods, and the results determined accuracy improved and real-time over other models. A neoteric framework proposed for detecting accident in Ijjina et al. [29]. For accurate object detection, Mask R-CNN capitalized in the proposed framework by an efficient centroid-based object tracking algorithm for surveillance footage. The basic idea is to determine an accident after overlapping vehicles together are speed and trajectory anomalies in a vehicle after an overlap with other vehicles. This framework was found to be dominant and paves the way to the development of general-purpose vehicular accident detection algorithms in real-time. The framework tested and evaluated by the proposed dataset with the different weather condition. In Saini et al. [30], a vehicle tracking technique based on image processing is developed without applying background subtraction for extracting the ROI. Instead, a hybrid of feature detection and region matching approach is suggested in their study, which is helpful for estimating the trajectory of vehicles over consequent frames. Later, as the vehicle path through an intersection, the tracked direction is monitored for the occurrence of any specific event. It is found that the proposed method has capability to detect an accident between two vehicles. Wenqi et al. [31] proposed the TAP-CNN model for predicting accident based on CNN in the highways. Traffic state and CNN model are described by some accident factors such as traffic flow, weather, and light to build a state matrix. In addition, the way of increasing TAP-CNN model accuracy for predicting traffic accident different iterations are analyzed. Accident data collected for inflected learning and evaluation of the model. Finally, the experimental results show that the proposed model Manguri and Mohammed: Traffic Controlling and Monitoring: A Review UHD Journal of Science and Technology | Jan 2023 | Vol 7 | Issue 2 11 named TAP-CNN is more effective than the traditional neural network model for producing traffic accidents. Dogru and Subasi [32] presented an intelligent system for accident detection in which vehicles exchange their microscopic vehicle variables with each other. Based on the vehicle speed and coordinates, data collected from vehicular ad-hoc networks (VANETs) simulated model in the proposed system and then, it sends traffic alerts to the drivers. Furthermore, it shows how to use machine learning methods to detect accidents on freeways in ITS. Two parameterizes help to analyze and detect accident easy which are position and velocity values of every vehicle. In addition, OOB data set has been used to test the proposed method. Finally, the results show that the RF is better than ANN and SVM algorithms by with 91.56%, 88.71, and 90.02% accuracy, respectively. Vision-based algorithms have been used in Yu et al. [33] to detect traffic accident including an ST-IHT algorithm for improving the robustness and sparsity of spatiotemporal features and weighted extreme learning machine detector for distinguishing between traffic accident and normal traffic. Furthermore, a two-point search technique is proposed to find a candidate value adaptively for Lipschitz coefficients to improve the tuning precision. For testing and evaluating the proposed method 30 traffic videos collected from YouTube website. Finally, the results show that the proposed method performance for detecting a traffic accident outperforms other existing methods. Accelerometer is a widely employed method that used to detect a crash. In this research work [34], after calibration of accelerometer value of acceleration is use to detect an accident. Due to the limitation of accelerometer accuracy and providing the efficient accident detection, CNN machine learning algorithm is tuned. For detecting an accident, image classification technique is used; however, CNN takes a lot of time, data, and computing power to train. Transfer learning methods have been innovatively applied to alleviate these problems and for the accident detection application, which involves retraining the already trained network. For this purpose, Inception-v3 classifier that developed by Google for image was incorporated. Finally, the proposed method efficiency compared with the traditional accelerometer-based techniques for detecting accident by 84.5% of accuracy for transfer learning algorithm. 4.4. Emergency Vehicles Detection The success of law enforcement and public safety relies directly on the time needed for first responders to arrive at the emergency scene. Emergency cars include ambulance, firefighter, and police car. Many methods are proposed to detect emergency cars and some of them as example are reviewed in this section. In Borisagar et al. [35], two methods of computer vision are used to detect and localize the emergency vehicle. The used methods including object detection and instance segmentation. The proposed method implementation includes Faster RCNN for object detection and Mask RCNN for instance segmentation. The results show that the proposed method is accurate, most importantly, and suitable for emergency vehicle detection in disordered traffic conditions are deliberated. In addition, a custom dataset used for detecting emergency vehicles which contains 400 images and labeled using the label me tool. Roy and Rahman [36] are proposed a model for detecting emergency cars from CCTV footage such as ambulance and fire-fighter on a heavy traffic road. In this model, priority given to these cars and clearing the emergency lane to pass the traffic intersection. For traffic police, sometimes deciding opening the specific lane for emergency vehicles is difficult or even impossible. Deep CNN and COCO dataset have been used for automated emergency car detection. The result of presented method for detecting and identifying all kinds’ emergency vehicles generally is reasonable. E. P. Jonnadula and P. M. Khilar [37] are presented a hybrid architecture for detecting emergency vehicles by combining features of image processing and computer vision. Also, search space decreased by using region of interest. Prediction of ambulance helps to decrease the number casualty on the real traffic in case of having emergency situation on the road. To cover this problem, Lin et al. [38] presented a novel approach-based machine learning techniques and features are extracted using the multi-nature which can extract ambulance characteristics on demand. Furthermore, they experimentally evaluate the performance of next-day demand prediction across several state-of-the- art machine learning techniques and ambulance demand prediction methods, using real-world ambulatory and demographical datasets obtained from Singapore. Finally, various machine learning techniques used for different natures and SCDF-Engineered-Socio dataset have been used to show the proposed method accuracy. The existing traffic light system is a lack of information in emergencies such as ambulances, firefighters, and police when a car is in. Suhaimy et al. [39] developed an embedded Manguri and Mohammed: Traffic Controlling and Monitoring: A Review 12 UHD Journal of Science and Technology | Jan 2023 | Vol 7 | Issue 2 Table 2: Proposed methods in literature reviews Type of traffic management Reference (s) Algorithm (s) Dataset (s) Accuracy % Contribution (s) Density Estimation [17] Block Variance The TrafficDB 93.70 Traffic density estimation with the low computational cost [1] Background Subtraction, over feat framework, and place meter ImageNet 96.55 Defining ROI by over feat framework [3] Detection (SSD) and MobileNet-SSD Data collected from cameras with different places 92.97 (SSD), 79.30 (MobileNet-SSD) New path opened for real time traffic density estimation [18] YOLO and DeepSORT Collected data from CCTV 87,88 (Day, Congestion) 93,88 (Day, Normal) 82,1 (Night, Normal) Detecting, tracking and counting vehicles [19] ROI and edge detection - N/A New technique developed for estimating traffic density [20] Background subtraction and morphological techniques - N/A Traffic density estimated [40] CNN UCSD 99.01 Traffic density estimation model proposed based CNN and computer vision Traffic Sign Detection and Recognition Approaches [22] HSI, HOG, LSS, and SVM GTSDB, CTSD 98.24 (GTSDB), 98.77 (CTSD) Developed Circle detection algorithm and an RGB-based color thresholding technique [21] HOG, LSS, Random Forest, and SVM GTSDB 97.04 In the first step, machine learning algorithms not used classify shapes instead of this invariant geometric moments have been used. Second, method has been proposed for the recognition [23] ROI, HOG, SVM, and Context Aware Filter GTSDB 99.43 (Prohibitory) 95.01 (Mandatory) 97.22 (Danger) Online detecting mandatory, prohibitory and danger traffic signs [24] Aggregate Channel Features and Boosted Trees Classifier Daimler, LISA and LaRA 84.314 (Daimler), 90.33 (LISA), 92.048 (LaRA) Proposed the high average-recall and speed method [26] HOG, LSS, and SVM GTSRB, GTSDB and TST 97.43 Shapes classified byusing invariant geometric moments [25] SGW and SVM GTSDB and CTSD 99.43 Speed of detection and classification improved which is more than 6 frames per second [27] CNN GTSRB 99.94 CNN process described [41] Proposed model named CapsNet TL_Dataset The proposed CapsNet is employed for traffic sign recognition. Accident Detection [28] Deep neural network model YOLO-CA CAD-CVIS 90.02 CAD-CVIS dataset created and the proposed method more fast and accurate [29] Mask R-CNN Proposed 71 Developing vehicular accident detection algorithms in real-time [30] Hybrid of feature detection and Region matching Real world dataset N/A Accident detection between two vehicles [31] CNN Accident data collected 78.5 Accident predicted using CNN [32] ANN, SVM, and Random Forests (RF) OOB data set 91.56 (RF), 88.71 (ANN), 90.02 (SVM) The proposed method can provide estimated geographical location of the possible accident (Contd...) Manguri and Mohammed: Traffic Controlling and Monitoring: A Review UHD Journal of Science and Technology | Jan 2023 | Vol 7 | Issue 2 13 machine learning application, including acquisition of data, features extraction, different algorithms exploration, tuning, and deploying the model to a good output model in a simulation application. Specifically, a classifier of ambulance siren sound into “Ambulance Arrive” and “No Ambulance Arrive” has been developed, which is the traffic light system could be used to track an ambulance’s arrival in an emergency. This paper suggests an approach based on Mel-frequency spectral Coefficients-SVM (MFCC-SVM) on MATLAB R2017b tools. 5. DISCUSSION According to the results of this review, several attempts have been made to develop intelligent traffic controlling vision- based methods. Some challenges can be seen when researchers try to develop vision based automatic traffic signals. One of the challenges is that there is no available framework to cover all traffic problems because huge amount of data and computational time are required. Another challenge is the power consumption to get real traffic data to testing their proposed methods in the different weather conditions. The Table 2: (Continued) Type of traffic management Reference (s) Algorithm (s) Dataset (s) Accuracy % Contribution (s) [33] ST-IHT, Spatio-Temporal Features and W-ELM Collected dataset 87.4±0.3 (SVM), 94.3±0.2 (ELM), 95.5±0.3 (W-ELM) (i) Robust Fractures extraction proposed based on OF-DSIFT and ST-IHT (ii) detect imbalance between traffic accident and normal traffic [42] YOLOv4 video sequences collected from YouTube N/A presents a new efficient framework for accident detection Emergency Vehicles Detection [35] Faster RCNN and Mask RCNN Custom dataset 81 (Object Detection), 92 (Instance Segmentation) The computational and accuracy for emergency vehicle detection are suitable [36] Deep convolutional neural network COCO 97.97 Detecting and identifying all kinds emergency cars [37] YOLO + ResNet COCO N/A Hybrid architecture presented for detection of emergency vehicles in a real time [38] SVR, MLP, RBFN, and LightGBM SCDF- Engineered-Socio N/A Varying degrees to the model training in LightGBM [39] MFCC-SVM - 97 Effectively distinguish audio events from audio signals SSD: Single Shot Detection, CNN: Convolution neural networks, HOG: Histogram of oriented gradients, LSS: Local self‑similarity, SVM: Support vector machine, GTSDB: German Traffic Sign Detection Benchmark, and GTSRB: German Traffic Sign Recognition Benchmark Table 3: Used datasets Type of Traffic Management Used in References Dataset (s) Type No. of Images Density estimation [17] The TrafficDB Image - [1] ImageNet Image 3.2 million [3], [18] Data collected from cameras with different places - [40] UCSD Video - Traffic sign detection and recognition approaches [21], [22], [23], [25], [26] GTSDB Image 50, 000 [24] Daimler Image 5,000 [24] LISA Video - [24] LaRA Video - [26], [27] GTSRB Image 900 [41] TL_Dataset Image 46,000 Accident detection [28] CAD-CVIS Video - [29], [30], [31], [33], [42] Proposed, real, and collected data - [42] YOLOv4 Emergency vehicles detection [35], [43] Custom dataset - [36], [37] COCO 328,000 GTSDB: German Traffic Sign Detection Benchmark, GTSRB: German Traffic Sign Recognition Benchmark Manguri and Mohammed: Traffic Controlling and Monitoring: A Review 14 UHD Journal of Science and Technology | Jan 2023 | Vol 7 | Issue 2 third challenge is the lack of standardized dataset for testing and training methods. The results show that there is no any comprehensive dataset for traffic controlling and monitoring. For example, in density estimation, most of the researchers have created their own datasets while in traffic sign detection and recognition; they have used some publicly available datasets such as (GTSDB and GTSRB). On the otherhand, both the accident and emergency vehicle detection methods have only collected and prepared (customized) real data which captured by CCTV cameras. Finally, current systems while studied in the literature provide a low-cost solution for traffic applications in the expense of the system accuracy and they are applicable. 5.1. Survey of Technique’s Summary According to the literature sur vey, researchers have proposed and developed many approaches for controlling and monitoring signal system based on computer vision algorithms. In Table 2, presenting methods for each survey topics associated with a summary of their (methods, datasets, and contributions). The reason that why the performance of the reviewed methods is not evaluated is a non-availability of a common datasets. 5.2. Datasets For testing and evaluating the proposed methods, researchers worked on the public datasets and on their collected datasets. The used datasets are described in Table 3. 6. CONCLUSION In recent years, reducing road congestion has become a key challenge because of the threatening rise in the number of vehicles on the roads. In this review, the existing studies on autonomic traffic controlling and monitoring are reviewed in the computer vision community. Furthermore, computer vision is considered as the areas with the most studies are for the future technologies. The intelligent traffic systems perceive the density estimation investigate, traffic sign detection and recognition, accident detection, and emergency vehicle detection. Furthermore, name of the used datasets in the reviewed papers are presented. The main gap that founded in this review is a non-availability of dataset for traffic controlling and monitoring. Finally, intelligent traffic systems can play a key role in reducing congestion in the intersection areas and traffic flow management. The conducted survey indicates the accuracy finding of each method as described in Table 2. This research work could be having a potential impact for further researches in the same field of study. Various challenges such as (weather conditions, lighting, and traffic patterns) can be considered with all techniques based on computer vision and machine learning methods. Consequently, these conditions will improve our survey in the future work. REFERENCES [1] D. Biswas, H. Su, C. Wang, J. Blankenship and A. J. S. Stevanovic. “An automatic car counting system using OverFeat framework,” Sensors, vol. 17, no. 7, p. 1535, 2017. [2] N. K. Jain, R. K. Saini and P. Mittal. “A review on traffic monitoring system techniques,” in Soft Computing: Theories and Applications, Singapore, Springer Singapore, pp. 569-577, 2019. [3] D. Biswas, H. Su, C. Wang, A. Stevanovic and W. Wang. “An automatic traffic density estimation using Single Shot Detection (SSD) and MobileNet-SSD,” Physics and Chemistry of the Earth, Parts A/B/C, vol. 110, pp. 176-184, 2019. [4] M. C. Coelho, T. L. Farias and N. M. Rouphail. “Impact of speed control traffic signals on pollutant emissions,” Transportation Research Part D: Transport and Environment, vol. 10, no. 4, pp. 323-340, 2005. [5] Q. Guo, L. Li and X. Ban. “Urban traffic signal control with connected and automated vehicles: A survey,” Transportation Research Part C: Emerging Technologies, vol. 101, pp. 313-334, 2019. [6] S. K. Kumaran, S. Mohapatra, D. P. Dogra, P. P. Roy and B. G. Kim. “Computer vision-guided intelligent traffic signaling for isolated intersections,” Expert Systems with Applications, vol. 134, pp. 267-278, 2019. [7] M. H. Malhi, M. H. Aslam, F. Saeed, O. Javed and M. Fraz. “Vision based intelligent traffic management system,” in 2011 Frontiers of Information Technology. IEEE, New Jersey, pp. 137-141, 2011. [8] C. J. Lakshmi and S. Kalpana. “Intelligent traffic signaling system,” in 2017 International Conference on Inventive Communication and Computational Technologies (ICICCT), pp. 247-251, 2017. [9] P. Jing, H. Huang and L. J. I. Chen. “An adaptive traffic signal control in a connected vehicle environment: A systematic review,” Information, vol. 8, no. 3, p. 101, 2017. [10] S. S. S. M. Qadri, M. A. Gökçe and E. Öner. “State-of-art review of traffic signal control methods: challenges and opportunities,” European Transport Research Review, vol. 12, no. 1, p. 55, 2020. [11] B. Ghazal, K. ElKhatib, K. Chahine and M. Kherfan. “Smart Traffic Light Control System,” In: 2016 3rd International Conference on Electrical, Electronics, Computer Engineering and Their Applications (EECEA), IEEE, United States, 2016, pp. 140-145. [12] H. Jeon, J. Lee and K. J. Sohn. “Artificial intelligence for traffic signal control based solely on video images,” IEEE Acces, vol. 22, no. 5, pp. 433-445, 2018. [13] Y. Wang, X. Yang, H. Liang and Y. Liu. “A review of the self-adaptive traffic signal control system based on future traffic environment,” Journal of Advanced Transportation, vol. 2018, 1096123, 2018. [14] S. D. Khan and H. Ullah. “A survey of advances in vision-based vehicle re-identification,” Computer Vision and Image Understanding, vol. 182, pp. 50-63, 2019. [15] Y. S. Huang and T. H. Chung. “Modeling and analysis of urban traffic light control systems,” Journal of the Chinese Institute of Engineers, vol. 32, no. 1, pp. 85-95, 2009. [16] K. H. K. Manguri. “Traffıc Sıgnalıng Control at Hıghway Manguri and Mohammed: Traffic Controlling and Monitoring: A Review UHD Journal of Science and Technology | Jan 2023 | Vol 7 | Issue 2 15 Intersectıons Usıng Morphologıcal Image Processıng Technıque,” Türkiye, Hasan Kalyoncu Üniversitesi, 2016. [17] K. Garg, S. K. Lam, T. Srikanthan and V. Agarwal. “Real-time road traffic density estimation using block variance,” in 2016 IEEE Winter Conference on Applications of Computer Vision (WACV), New Jersey, IEEE pp. 1-9, 2016. [18] K. H. N. Bui, H. Yi, H. Jung and J. Cho. “Video-based traffic flow analysis for turning volume estimation at signalized intersections,” in Intelligent Information and Database Systems, Cham, Springer International Publishing, pp. 152-162, 2020. [19] F. Kurniawan, H. Sajati, O. Dinaryanto. “Image processing technique for traffic density estimation,” International Journal of Engineering and Technology, vol. 9, no. 2, pp. 1496-1503, 2017. [20] B. Eamthanakul, M. Ketcham and N. Chumuang. “The traffic congestion investigating system by image processing from cctv camera,” in 2017 International Conference on Digital Arts, Media and Technology (ICDAMT), New Jersey, IEEE, pp. 240-245, 2017. [21] S. K. Berkaya, H. Gunduz, O. Ozsen, C. Akinlar and S. Gunal. “On circular traffic sign detection and recognition,” Expert Systems with Applications, vol. 48, pp. 67-75, 2016. [22] Y. Yang, H. Luo, H. Xu and F. Wu. “Towards real-time traffic sign detection and classification,” IEEE Transactions on Intelligent Transportation Systems, vol. 17, no. 7, pp. 2022-2031, 2016. [23] S. Salti, A. Petrelli, F. Tombari, N. Fioraio and L. Di Stefano. “Traffic sign detection via interest region extraction,” Pattern Recognition, vol. 48, no. 4, pp. 1039-1049, 2015. [24] X. Du, Y. Li, Y. Guo and H. Xiong. “Vision-Based Traffic Light Detection for Intelligent Vehicles,” in 2017 4th International Conference on Information Science and Control Engineering (ICISCE), pp. 1323-1326, 2017. [25] F. Shao, X. Wang, F. Meng, T. Rui, D. Wang and J. J. S. Tang. “Real-time traffic sign detection and recognition method based on simplified Gabor wavelets and CNNs,” Sensors, vol. 18, no. 10, p. 3192, 2018. [26] A. Ellahyani, M. E. Ansari and I. E. Jaafari. “Traffic sign detection and recognition based on random forests,” Applied Soft Computing, vol. 46, pp. 805-815, 2016. [27] A. Shustanov and P. Yakimov. “CNN design for real-time traffic sign recognition,” Procedia Engineering, vol. 201, pp. 718-725, 2017. [28] D. Tian, C. Zhang, X. Duan and X. Wang. “An automatic car accident detection method based on cooperative vehicle infrastructure systems,” IEEE Access, vol. 7, pp. 127453-127463, 2019. [29] E. P. Ijjina, D. Chand, S. Gupta and K. Goutham. “Computer Vision-based Accident Detection in Traffic Surveillance,” in 2019 10th International Conference on Computing, Communication and Networking Technologies (ICCCNT), pp. 1-6, 2019. [30] A. Saini, S. Suregaonkar, N. Gupta, V. Karar and S. Poddar. “Region and feature matching based vehicle tracking for accident detection,” in 2017 Tenth International Conference on Contemporary Computing (IC3), pp. 1-6, 2017. [31] L. Wenqi, L. Dongyu and Y. Menghua. “A model of traffic accident prediction based on convolutional neural network,” in 2017 2nd IEEE International Conference on Intelligent Transportation Engineering (ICITE), pp. 198-202, 2017. [32] N. Dogru and A. Subasi. “Traffic accident detection using random forest classifier,” in 2018 15th Learning and Technology Conference (L&T), pp. 40-45, 2018. [33] Y. Yu, M. Xu and J. Gu, “Vision-based traffic accident detection using sparse spatio-temporal features and weighted extreme learning machine,” IET Intelligent Transport Systems, vol. 13, no. 9, pp. 1417-1428, 2019. [34] P. Borisagar, Y. Agrawal and R. Parekh. “Efficient vehicle accident detection system using tensorflow and transfer learning,” in 2018 International Conference on Networking, Embedded and Wireless Systems (ICNEWS), pp. 1-6, 2018. [35] S. Kaushik, A. Raman and K. V. S. R. Rao. “Leveraging computer vision for emergency vehicle detection-implementation and analysis,” in 2020 11th International Conference on Computing, Communication and Networking Technologies (ICCCNT), pp. 1-6, 2020. [36] S. Roy and M. S. Rahman. “Emergency vehicle detection on heavy traffic road from CCTV footage using deep convolutional neural network,” in 2019 International Conference on Electrical, Computer and Communication Engineering (ECCE), pp. 1-6, 2019. [37] E. P. Jonnadula and P. M. Khilar. “A New Hybrid Architecture for Real-Time Detection of Emergency Vehicles.” In: Computer Vision and Image Processing. Springer, Singapore, 2020, pp. 413-422. [38] A. X. Lin, A. F. W. Ho, K. H. Cheong, Z. Li, W. Cai, M. L. Chee, Y. Y. Ng, X. Xiao and M. E. H. Ong. “Leveraging machine learning techniques and engineering of multi-nature features for national daily regional ambulance demand prediction,” International Journal of Environmental Research and Public Health, vol. 17, no. 11, p. 4179, 2020. [39] M. A. Suhaimy, I. S. A. Halim, S. L. M. Hassan and A. Saparon. “Classification of ambulance siren sound with MFCC-SVM,” in AIP Conference Proceedings, United States, AIP Publishing LLC vol. 2306, no. 1, p. 020032, 2020. [40] L. A. T. Nguyen and T. X. Ha. “A novel approach of traffic density estimation using CNNs and computer vision,” European Journal of Electrical Engineering and Computer Science, vol. 5, no. 4, pp. 80- 84, 2021. [41] X. Liu and W. Q. Yan. “Traffic-light sign recognition using Capsule network,” Multimedia Tools and Applications, vol. 80, no. 10, pp. 15161-15171, 2021. [42] H. Ghahremannezhad, H. Shi and C. Liu. “Real-time accident detection in traffic surveillance using deep learning,” in 2022 IEEE International Conference on Imaging Systems and Techniques (IST), pp. 1-6, 2022.