Reserach Project Scope

Literature Survey

The existing parking slot detection systems are categorized into two types, based on the methods used to identify parking slots: camera based systems and sensor based systems. Some of the systems focused on images rely on computer vision techniques to detect parking slots. In this paper denotes the computer vision techniques and mainly canny edge detection, to classify the model as occupied or available.

Methods such as support vector machines and texture classifiers have also been used to design parking management systems in existing systems. However, these methods rely on extensive, manual parking slot labelling a single. Edwin K Jose, et al. created a vehicle occupancy detection framework using CNN-based YOLO architecture along with the Kalman filter tracking algorithm for the transient occlusion handling. Most current location prediction systems are typically based on the global satellite navigation system (GNSS), such as the global positioning system (GPS).

Some research and technological innovations have enhanced the accuracy of the position equipment based on satellites. However, a single GNSS quality couldn’t compensate for accurate, efficient and continuous positioning in signal-degraded environments [6].Several methods to estimate the position use additional devices and wireless sensors. Dead-reckoning using extra sensors such as Inertial Measuring Unit (IMU) and speed measurements have been used to boost positioning accuracy and availability . Furthermore, an integrated navigation system combining GPS and IMU, a system by detecting lane using digital images and lateral lane distance information, matching the threedimensional (3D) point cloud results produced during driving by Light Detection And Ranging(LIDAR) and matching the vision and radar sensor-based road geometry information . Such approaches require separate network connectivity or additional infrastructure and sensors; accordingly, drawbacks of requiring a significant amount of measurement and expense were weighed.

Reserach Gap

Research Target Mobile/Web application Drawbacks Technologies
Smart Parking System (SPS) architecture using ultrasonic detector Smart Parking System (SPS) architecture using ultrasonic detector Used the Sensor Based method. One sensor used for each parking slot. Ultrasonic sensors are used for parking slots detection and improper parking detection. Mereological conditions for lightning variations are low Sensors technologies. Mereological conditions for lightning variations are low Sensors technologies
Error Correction Method with Precise Map Data for GPS/DR Based on Vision/Vehicle Speed A ready-made speed sensor from ECU in the vehicle and equipped vision sensor System used and Using a vehicle speed sensor is not Applicable for every user Vision Computing
Review-Recognition of License Number Plate using Character Segmentation and OCR with Template Matching This approach has been improved to partition and partition all letters and numbers as part of the license plate using border-box technology. Low sensitive and able to locate the plate at any conditions Character Segmentation, OCR , Template Matching
Car Park Management with Networked Wireless Sensors and Active RFID Combines readers and RFID tags sensors for detection of parking spaces Cost is high for a large parking area RFID technology
AI-Based Targeted Advertising System Model for recognizing various types of objects are trained separately by different architectures. High computational complexity And the cost is high Machine learning technology
Image recognition and processing using Artificial Neural Network Provide a new approach for image recognition using Artificial Neural Networks. Initially an original gray scale intensity image has been taken for transformation. Scalability is low and no proper system Soft computing models on digital image, Artificial Neural Networks
Error correction with PPP technique Test results were also faced with the estimate accuracies for four PPP measurement models. High computational complexity PPP, UD, SD, TD, and TSD models
Proposed System “iParking” – Smart way to automate the management of the Parking System for a Smart City Propose a proper management for outdoor parking parking slots availability and occupancy detection with real time environment positioning accuracy enhancement approach for global positioning system. Identify a vehicle’s license plate and activate access control systems for automatic gate access to authorized members. In order to build not only accurate but also robust detection model, suggest that it is important to account for cars that are moving and not parked yet, unauthorized parked cars. These maybe further steps to investigate in the future. Image Processing, Map Matching,Machine Learning

Reserach Problems

In many metropolises, the parking problems are becoming increasingly important. Following the rapid enhancement of traffic demand, the imbalance between parking supply and parking demand has been considered as the main reason for metropolis parking problems.

In outdoor vehicle parks searching for unoccupied parking slots and not having accurate GPS root guidance towards the parking area is time consuming for the user.

Customers need a reliable and user-friendly payment solution according to their entry/exit time

Reserach Objectives

The main objective of this research is to design and implement a smart-parking system will allow drivers to check for available parking spaces beforehand and provide root navigation and GPS error correction towards the parking area. Users are able to pay for the parking fee for exact time they have used. This system also provides promotions and sales information for a target audience.



Error analysis for Global Positioning System (GPS)

Although GPS signals are available most of the time, a location may be invalid sometimes due to Satellite signal blockage due to buildings, bridges, trees or the reception of receivers therefore finding an accurate positioning is still challenging for supporting the vehicle navigation. These errors in positioning can lead to incorrect navigation and time wastage. Map matching is needed to accurately locate a vehicle on a given road segment to result in the vehicle location accuracy. ... The points obtained from the roads before the algorithm is run are used as map data. MM is used to give a proper section of the road to form a lane. Path nodes are denoted by setting the actual location longitude and latitude. The proposed method uses the following point sets
• When a vehicle moves along the road, the map point set is constructed to execute the coordinates extracted from the center of the roads.
• Vehicle position source coordinates are updated from a GPS receiver with the location points for the vehicle.
• Set of reference points is generated from the actual coordinates chosen in the map where a vehicle is supposed to travel.

Main Steps
 Computation of the X, Y, and Z coordinates of a given point with its latitude, longitude to get the vehicle coordinates.
 Setting of real road co-ordinates
 Finding the triangle between the three points.
 Item Calculates the distance of two coordinates.
 Returning the matched position for two coordinates and vehicle coordinate.
 Executing the GPS data for the matched position. On the foot of the triangle.


Checking slots availability in parking

Available parking slot detection using a classification approach which can deal with outdoor parking. Such a system is highly useful in scenarios where a camera is placed at a lamp post view and parking slots are visible. It returns the real time states of the parking slots providing the number of available slots and its specific positions in order to guide the drivers through the roads. This research will implement based on the theories of Background Subtraction Algorithm to mapping methods of reducing the error of vehicle detection and ... vehicle detection and motion tracking to distinguish between the occupied and available slots. Finally, the status of parking slots is offered for users through an android app with real-time feedings. One of the keys to the identification of a movable vehicle is the background subtraction algorithm, and it was one of the layers of the main program that operated as a filter to eliminate every error in detection when evaluating image frames. To be able to identify the vehicle continuously always, particularly when the vehicle was in static or dynamic, the software needs to learn the image of the vehicle. That's when it comes to machine learning. Haar Cascade Classifier is using to Object Detection in this proposed system, this appropriate machine learning technique that is widely available for use in image processing is referred to as a cascaded classifier. This system is proposed, which classifies the parking slot as occupied or available. When distinguish between the slots occupied and available for system uses a combination of Laplacian operator to detect edges in vehicles and HAAR Cascade classifier for vehicle identification and motion tracking. The background subtraction algorithm, morphological operations, and contours and are used for moving vehicle detection. Further, Pedestrian detection using HOG and SVM is incorporated to detect pedestrians in the parking region. The Laplacian operator is used to enhance the accuracy of classification as available or occupied slot. A Laplacian threshold value is established because the possibility of the presence of is maximum vehicles. Basically, parking regions are laid out manually and a threshold is defined by calculating the magnitude of edges inside of them. The vehicle will have multiple noticeable edges, while the free parking slots are smooth. This is the fact behind the use of the technique of Laplacian. If the probability of a vehicle being present is above the threshold value, the classifier shall be used to detect whether it is a vehicle. Whenever a difference in the threshold value is detected, the classifier is called to verify whether that is a vehicle. It is found to be a vehicle, so it is noted and observed for motion. Whenever there is a motion at the stage, the classifier is called again to evaluate the existence of the vehicle. The device therefore recommends a combination of identification, classification and monitoring in order to ensure effective parking.


A Real-Time license plate detection system for parking access

Automated parking approach using automated license plate identification (ALPR) and real-time embedding system for access control is common in urban spaces these days. this paper implements a method for the license plate detection system using OpenALPR open-source library. The main purpose of the system is to identify all vehicles entering and exiting through the barrier parking system and send an automated notification about the payment details to the ... owner of the relevant vehicle. When the vehicle reaches the barrier, the LP identification unit automatically identify the LP registration number and differentiate it with the user registration description. If there is a match, the entrance door will open. After the vehicle is parked, the LP registration number of the redesigned unit is automatically read near the exit door. after that send the notification to the user about payment details. In this proposed system, software framework consists of three parts. One to perform the ALPR function, one to provide gate control, and another to the notification system. In ALPR Function we used eight pipeline phases. They are Detection, Binarization, Character Analysis, Plate Edges, Deskew, Character Segmentation, OCR and Post Processing. Here we use two gates namely the entry gate and the exit gate. Upon vehicle exiting gate the parking lot, the system identifies the vehicle number plate using image processing technology, deals with the relevant database, Furthermore, opens the exit door and provides the relevant payment details to the customer.


Payment pattern, Reservation, and effectiveness of digital advertising

We introduce a parking pattern for the parking reservation system. Dataset groups into clusters using the K-Means clustering algorithm and k nearest neighbor algorithm. The method that allows grouping the hours into the different clusters whose centroid represents the count of the vehicle measures in the hours of the day. Dataset clustered into three clusters ... as cluster 0, cluster 1, cluster 2. Then Checking by the vehicle count in each cluster separates the clusters as peak clusters, non-peak clusters, and normal clusters. According to the clustering results, the peak cluster time (Hours) selected as peak hours, non-peak cluster time period (Hours) selected as Non-peak hours, and remaining hours selected as normal hours. Hours that are in the peak cluster rental are high, the non-peak clusters rental is low and other hours’ rental is normal.Once the user going to reserve the parking slot, new users should register to the system first. After login to the system user able to navigate the reservation page. On the reservation page user able to see new reservations and view reservations. When navigating to the new reservations page user can enter the hour and search. According to the hours load all the available slots. Available slots display with green color and reserved slots display as red color. The firebase database has all the slot numbers and some details of the user. If a slot is reserved, the status of the slot is ‘1’, and user details are updated. Otherwise, it is ‘0’. Once the user selects the green color button user navigates to the Slot reservation page in there, the user can see the selected slot number top of the page, and the user should add some details like vehicle number, User name, NIC number, Phone number, time which are going to reserve and total hours which the user going to reserve the slot. Then the system checks the time which the user entered. If the searched hour is in the peak hours, peak hour rental multiply with total hours and it is displayed in the rental field. If the searched hour is in the non-peak hours, non-peak hour rental multiplies with total hours and it is displayed in the rental field. If the searched hour is in the normal hours, normal hour rental multiply with total hours it is displayed and in the rental field. After click on the reserve button user navigates the summary page. On that page, the user can see a summary of the reservation. Then the user should pay the rental. After click on the Pay button user navigates to the rental page, on that page user should add some card details and click the confirm button. After completing the reservation step user receives a successful reservation message. When discussing the effectiveness of digital advertising it is cost-effective for the displayer and the shop owner both. First I calculate the count each advertisement displayed time at a specific period. Then calculate the cost of the displaying advertisement and charge from the shop owner to advertise. Then calculate the profit gain from the digital advertisement for the parking area owner


Technology Used