Description:REAL TIME VEHICLE DETECTION AND COUNTING FOR TRAFFIC MANAGEMENT APPLICATIONS
Field of Invention
The innovation relates to the use of machine learning algorithms to identify and count the vehicle detection and manage the traffic in an efficient manner.
The Objectives of this Invention
The main objective of the innovation is being proposed here is to vehicle count identifications as well as vehicle movement in the traffic can be predicted based on the input videos.
Background of the Invention
In recent years, the manufacturing of the vehicle is increased day by day, because of that traffic also increased lot. First, type of technique has been introduced in (US2006/7884739B2), Numerous sensors detect traffic disruptions and capture data on them, with the data reflecting both the entity involved in the trouble and the number of cars affected. A traffic infringement is suspected when enough suspicious data is collected, and a fine is calculated based on at least the percentage of vehicles involved. The offending party can then be notified by having a message delivered to its electronic device to collect the fee. The notification can include a link to an online payment system or a form to file an appeal. Furthermore, those who suffered due to the congestion may be singled out and compensated by the fine proceeds. Another method, (WO2021/077766A1), An infrastructure and approach to detecting traffic events across several targets in a vast area. A microwave detector sends a target's position to a video detector, which in turn detects feature data concerning targets, followed by a traffic analyzer, which matches the details recognized by the radio frequency detector with the knowledge identified by the video detection system to determine the exact location of moving targets on the roadway. By combining microwave and video, traffic regulation agencies can better control and regulate fast vehicle behavior within an acceleration limit region and illegal lane occupation. In (CN2011/102231231A), The invention describes an arrangement and technique for providing early warnings of potentially dangerous traffic situations on a regional road network. Monitoring equipment, sending and receiving data, processing that data, and issuing an early warning are all part of the system. Data on road traffic, weather, illegal behaviors, and collisions are collected using equipment for monitoring, transmitted using a data transmission system, subjected to multiple heterogeneous treatments in a data processing system, merged with data of the identical kind, analyzed, and graded using an early warning system. Finally, the cause of an incident is determined. In this way, we can get a complete picture of the road's traffic safety conditions and quickly identify any potential threats to assign a proper notification grade and implement effective countermeasures. The effectiveness, timeliness, and reliability of measuring whether or not a road traffic management agency is actively preventing traffic accidents have all been enhanced. The invention provides technological aid in reducing traffic risk factors on area road networks and preventing big and severe road traffic accidents. In (CN2019/111144945A), A particular version of the tool gives a server and an automobile passing administration system well-suited to the transportation sector. The method includes the following steps: receiving establishing knowledge of a vehicle obtained by a vehicle-mounted terminal, determining the driving path of the car on a target road based on the positioning information, and determining the vehicle's toll along the target road. The vehicle's driving course, and distance can be identified using positioning technology, and then the toll is collected utilizing a vehicle terminal positioning and cloud levying mode, allowing for precise pushing of the vehicle according to the actual driving road section; simultaneously, scrutiny and oversight of fee escape automobiles can be achieved through a combination of automatic recognition and real-time monitoring.
The (Zhang et al [2021], IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 1, pp. 404-415), Researchers and consumers alike have become interested in Intelligent Traffic Signal Control (ITSC) systems because of their potential to reduce gridlock. Vehicle to Infrastructure (V2I) wireless communications have recently offered a cost-effective method to achieve ITSC by detecting cars. However, a V2I solution would only pick up on vehicles with wireless communications capability, whereas most traditional ITSC algorithms assume that every vehicle is identified. Here, we look at the 'Partially Detected Intelligent Transportation Systems' subset of the transportation system family. Due to the gradually growing penetration rates of the underlying technologies, such as Dedicated Short-Range Communications (DSRC) technology, an algorithm that can function effectively under a tiny detection rate is hugely desirable. When only a subset of vehicles are picked up by the ITSC system, the reinforcement learning (RL) approach inside AI could supply crucial resources. In this research, we provide a novel RL algorithm for use in PD-ITSC systems, which monitor and adjust traffic lights based on only partially detected information. This system's effectiveness is examined in various scenarios with varying numbers of vehicles, detection rates, and road configurations. Even with a poor detection rate, our technology is nevertheless able to minimize the average waiting time of vehicles at a junction, which reduces the travel time of vehicles.
The (Zuraimi et al [2021], 2021 IEEE 11th IEEE Symposium on Computer Applications & Industrial Electronics (ISCAIE), Penang, Malaysia, 2021, pp. 23-29), This paper research uses TensorFlow's machine learning framework and the object detection algorithm you only look at once (yolo) to build deep learning for real-time vehicle recognition. The approach proposed in this study for determining the advancement of YOLOv4 newest algorithm over the prior model in vehicle detection system combines these two assumptions with python as a a coding language. The DeepSORT algorithm is also used for this vehicle detection to aid in precisely calculating the number of vehicles seen passing by in the video. According to this article, with state-of-the-art results of 82.08% AP50 on the custom dataset at a real-time speed of roughly 14 FPS on GTX 1660ti, Yolov4 is the best model among YOLO models.
Summary of the Invention
It also works on the Single way and Two Ways. It counts the vehicle in specific direction, it may be from upside to downside or downside to upside. To obtained all the results we have used background subtraction algorithm to locate the vehicles in each frame, and later we have detected the vehicles and incremented the count.
Detailed Description of the Invention
The technique may be used to find, identify, and track cars in video sequence, then divide the found cars into 3 variants according to respective dimensions. The system proposed consists of three functionalities: foundation acquisition, salient extracting, and transportation classification, as shown in the diagram. A common approach for getting the salient image or spotting moving items is backstory removal.

This is the beginning of the process. In this case, the foreground in a video sequence may be thought of as the flowing items, and the backdrop can be thought of as the artefacts. If we were to imagine a traffic film, the road would serve as the foreground, and the movement items, such as the cars, would serve as the forefront. The above- mentioned technique employs image processing techniques to learn about the background. When the background subtraction is done, we can view the foreground, moving objects like car or vehicles. foreground is obtained by making the static pixels to connect each other and forming a contour. Foreground plays a major role in the process. I define a ROI by drawing a tight line on the image in the first frame of the video. The idea is to recognize that ROI afterwards, but that ROI isn't a key vehicle. It is just a component of a vehicle that can bend, rotate, translate and even be out of the frame entirely A deep CNN framework (AttentionNet) developed an active technique to choose a search window for vehicle recognition using an image context to capture the vehicle by sequential operations with top-down attention. By sequentially refining the bounding boxes, AttentionNet was able to attain satisfactory results on the vehicle detection benchmark. Proposed a sequential search technique to detect visual vehicles in photos, with the detection model trained using a deep RL framework to choose the best action to capture a vehicle in an image. This module will count identified vehicles and update the counted results on a regular basis based on vehicle detection. The results will be printed as a streaming video using OpenCV. Following the definition of the ROI, the system executes a sequence of operations, including applying a background mask, deleting the mask, executing binary threshold, morphology with erosion and converting the frame to grey scale. Following these actions, contours are discovered. The attributes of contours are extracted after the estimated centroid is in the diagonal range. Finally, the output is used to increment the respective variables.
An ROI is a specific area of a picture where an operation will be conducted. Instead of editing the entire image, ROI allows you to work with just a portion of it. The selection of a region of interest is critical in the proposed approach to minimize false positives in vehicle identification and classification. After that, the user selects the four places on the video that define the region of interest with his mouse.
Background Subtraction is removing the background part from the frame and obtaining the foreground static objects like moving vehicles in a video. If we already have an image of the background, such as a building or a road, background subtraction may be simple. Background images can be deleted and foreground items can be obtained in the situations described above, however this is not always the case. Background information. Backgrounds can be dynamic, and initial scene information may not be available. Additionally, the suggested technique gets more challenging since the reflections will be recognised as neighbouring pixels also if the elements with in stream generate reflections since these motion with individuals or automobiles. For cases like this, several methods have been presented; Some of these algorithms are developed in video Sequences, including such Scenery Decrement MOG Godbehere et al (2012), that creates a representation of the article's backstory using Time series. This is accomplished by using 3 to 5 Gaussian distributions. According on Zivkovic (2004) and Zivkovic et al (2006), Background Multiplexers GMG implements thresholding technique in computer Vision and integrates Probabilistic classification with the background color assessment approach. Figure 3 shows a camera viewfinder with the backdrop cropped off both prior to and after. The methodology employed in the suggested program's development is MOG2.
Contour is linking the pixels of the background subtraction and forming a closed loop. Canny edge detection is used to find the accuracy. The cv2.findContours() method in OpenCV is used to locate contours. Check to see if the device's location has exceeded the ROI. Whenever two ROI points are joined horizontally, the hypothetical line is formed. The system counts the car once its centroid in ROI crosses the imaginary line.
Vehicle count and traffic management are two key aspects of transportation and urban planning that aim to optimize the flow of vehicles and ensure safe and efficient mobility in urban areas. Here's a brief explanation of each: Vehicle count refers to the process of quantifying the number of vehicles traveling on a specific road, street, or intersection during a defined period. This data is typically collected through various methods, including manual observations, video cameras, radar sensors, or automated vehicle counting systems. Vehicle count data is essential for several purposes: Traffic Planning: It helps transportation authorities and urban planners understand traffic patterns and make informed decisions about road design, capacity expansion, and infrastructure improvements. Congestion Monitoring: Vehicle count data can identify areas prone to traffic congestion, enabling authorities to implement strategies to alleviate bottlenecks and reduce delays. Safety Analysis: It aids in assessing the safety of roads by identifying high-accident locations, allowing for targeted safety measures. Environmental Impact Assessment: Vehicle count data is used to estimate emissions and assess the environmental impact of traffic in urban areas.
Traffic management involves the implementation of strategies and measures to regulate and control the movement of vehicles on road networks to ensure safe, efficient, and sustainable transportation. Some key elements of traffic management include: Traffic Signals: The timing and coordination of traffic signals at intersections to optimize traffic flow and reduce congestion. Lane Management: Designating lanes for specific purposes, such as carpool lanes, bus lanes, or bicycle lanes, to improve overall traffic efficiency. Speed Limits: Setting appropriate speed limits to enhance safety and reduce accidents. Road Signs and Markings: Installing clear and informative road signs, lane markings, and pavement markings to guide drivers and reduce confusion. Intelligent Transportation Systems (ITS): Utilizing technology, such as real-time traffic monitoring and information dissemination, to provide drivers with timely updates on road conditions and alternative routes. Public Transportation Integration: Promoting the use of public transit systems to reduce the number of private vehicles on the road and ease congestion. Congestion Pricing: Implementing tolls or fees during peak hours to manage demand and reduce traffic congestion. Efficient vehicle count and effective traffic management strategies are crucial for creating livable cities, reducing travel times, improving air quality, and enhancing overall quality of life for urban residents. These practices are continually evolving as technology advances and cities seek innovative solutions to address their transportation challenges.

4 Claims & 1 Figure
Brief description of Drawing
In the figure which are illustrate exemplary embodiments of the invention.
Figure 1, The Process of Proposed Invention , Claims:The scope of the invention is defined by the following claims:

Claim:
1. A system/method to identify the vehicle count and manage the traffic in an efficient manner using machine learning algorithms, said system/method comprising the steps of:
a) The system starts with datasets collection from various cameras (1), from the video analysis will start (2).
b) The system is incorporated with removing of the background of the data (3), analyzing the divided frames (4), to identify vehicle images (5), the algorithm will trigger to count the vehicles (6), the image is matched and accuracy metric was compared and produce the vehicle count output (7).
2. As mentioned in claim 1, the invented system starts with various videos and image dataset uploading to start the process.
3. According to claim 1, the preprocessing will initiate to remove the noisy data from the dataset and it will trigger feature extraction process to remove the background of the videos then convert into the frames.
4. According to claim 1, now, the proposed invention will start from be matched with captured the videos to count the vehicle as well as clear the traffic management efficiently.