Vehicle License Plates Recorded recognition is a problem that needs to solve in many artificial intelligent transportation application scenarios, such as car access control in residential areas, automatic timed toll collection in parking lots, and automatic identification and capture of road electronic eyes for violations. Many companies are contributing to this field by developing AI technology that tackles various aspects of license plate recognition. One example is SentiVeillance.
Here are the articles to explain, Application of artificial intelligence in Vehicle License Plates Recorded
This paper mainly expounds on the whole process of license plate recognition and related artificial intelligence algorithms compare and analyzes the characteristics, advantages, and disadvantages of various algorithms, and has certain reference significance for further research on the application of artificial intelligence technology in the field of intelligent transportation.
Typical, vehicle license plate recorded recognition process generally includes vehicle image acquisition, license plate location, license plate character segmentation, and license plate character recognition in sequence. This paper elaborates on these contents in detail, including the problems to solve, the implementation process, and the algorithms used, more alpr info here.
Vehicle Image Acquisition
The first step in identifying a license plate is to acquire a vehicle image, that is, the acquisition of a vehicle image. In the scene of license plate recognition, vehicle images stand generally captured by cameras in real time. And the working environment is relatively complex, as is affected by various factors. Such as light, weather, vehicle speed, license plate position, etc. Therefore, it is first necessary to ensure the collected images on the hardware The vehicle images are as clear, complete, consistent in size, and easy to handle as possible. There are three main methods of vehicle image acquisition:
- (1) Video recognition. Its working principle is to identify vehicles by shooting video with a license plate recognition camera. Which can be an ordinary camera + video capture card, or a digital camera.
- (2) Ground induction coil identification. Its working principle is that when the vehicle enters the recognition area, the ground induction coil installed in front of the vehicle speed bump detects the vehicle and then sends a signal to the camera for capturing. Its advantage is that it has a high trigger rate and its performance is relatively stable. Its disadvantage is that it needs to be The hardware equipment stands specially installed, and the project volume is relatively large.
- (3) Video + ground induction coil recognition. Its working principle is to recognize the vehicle through the video taken by the license plate recognition camera. Then trigger the output recognition result through the ground induction coil. Compared with using ground induction coil recognition, this method has a faster recognition speed and higher recognition rate.
License plate positioning
The task of license plate location is to find and intercept the license plate area from the vehicle image collected in the previous step. License plate location algorithms can summarize into two categories:
(1) Location methods based on graphics and imaging, such as edge detection location, color location, etc. The interference of external information
will deceive the positioning algorithm, causing the positioning algorithm to generate too many non-license plate candidate areas and increase the system load. The implementation of this algorithm is
divided into three steps:
- Preprocessing the image such as edge detection and binarization;
- Performing morphology;
- Finding the license plate outline and positioning it accurately.
(2) Positioning methods based on machine learning, such as feature engineering positioning, neural network positioning, etc. The key to using this algorithm
for license plate location is to find good features and training methods. The main steps can divide into;
- Providing a set of training data (vehicle images) with correct output;
- Constructing a neural network model;
- Using the trained model to perform actual license plate location detection and location effects.
License plate character segmentation
The task of vehicle license plate recorded character segmentation is to correctly intercept the characters in the license plate area obtained in the previous step, and become multiple images containing only one character. Commonly used character segmentation algorithms include algorithms based on connected domain marking, algorithms based on character geometric features, and algorithms based on image projection. Among them, the projection algorithm stands more widely used, and the separation efficiency is higher. The implementation can describe as;
- First, traverse the entire image to calculate the number of white pixels (license plate number area) in each column and store them in the array;
- Secondly, obtain the corresponding projection image from the gray value in the array, and then pass. The content of the array finds the split point between adjacent characters;
- Finally, character segmentation stands done according to the split point.
License plate character recognition
License plate character recognition is the last step in the vehicle license plate recorded recognition process. The main task is to separate the license plate characters separated in the previous step. The images are converted into the correct characters and finally stitched into the correct license plate number. The actual application scenario mainly involves two aspects of recognition accuracy and recognition speed. In general, the recognition rate of the parking lot should reach more. Then 95% when the license plate is not damaged, and the recognition speed should reach within 1 second. License plate character recognition algorithms can be simply divided into the following three categories.
(1) Template matching.
This is an early and more traditional method of character recognition. The principle is to first establish a standard character template library for all possible strings contained in the license plate. Then process a single-character image into the same format as the characters in the template library. Finally, compare it with the characters in the template library according to certain rules. A similarity value calculates, and the character with the largest value is the correct character. The algorithm has a higher recognition rate for clear, clean, non-slanted, and deformed license plate character images.
(2) Neural network.
It is a machine learning algorithm, such as a typical BP neural network, or convolutional neural network (CNN). Its working principle is to first extract character features. Such as gradient distribution features and gray-level statistical features. Then build a neural network model and set network training parameters. Finally, perform model training and recognition inspection.
(3) Support Vector Machine (SVM).
The algorithm is essentially similar to the neural network algorithm, and it is also a machine learning method. Its main design idea is: first obtain the sample features, perform training, and then classify. The mathematical knowledge involved in the principle of SVM is relatively complicated. It is difficult to program and realize it by yourself. In practical applications, the more mature toolbox that supports. The SVM algorithm stands currently widely used.
What is the application and prospect of artificial intelligence in license plate recognition?
License plate recognition is a problem that needs to solve in many intelligent transportation application scenarios. Such as car access control in residential areas, and automatic timing toll collection in parking lots. Automatic identification and capture of violations by road electronic eyes. In recent years, artificial intelligence technology has stood applied to many vehicle license plates recorded recognition schemes.
This paper mainly expounds on the whole process of license plate recognition and compares the relevant algorithms of artificial intelligence, and analyzes the characteristics, advantages, and disadvantages of various algorithms, which has certain reference significance for further research on the application of artificial intelligence technology in the field of intelligent transportation.
The use of license plate recognition technology will greatly reduce traffic violations and bad safety traffic accidents, also provide strong evidence for the post-processing of various traffic accidents and life and property safety and play an important role in my country’s traffic safety and other aspects. No matter what trigger method uses, a mature license plate recognition system can effectively monitor passing vehicles in real-time, and analyze and obtain various information such as license plate number, license plate color, and vehicle type. It provides strong security support for the security department to effectively combat theft, blocklist motor vehicles, check traffic hit-and-run vehicles, analyze traffic conditions, and increase public security management.
The intelligent transportation system based on license plate recognition can timely prevent the increasingly rampant motor vehicle anti-theft, anti-theft, fake license plates, security, security, black market transactions, and other security activities. Through the “electronic license plate” information installed and registered by the motor vehicle, the monitoring center can effectively remotely control, and grasp the image, digital information, and driving direction of the suspicious vehicle, and feedback on the tracking information to the monitoring center at any time.
Based on this information, security departments can keep abreast of, track, and control illegal vehicle transactions, vehicle theft, and other security behaviors. When the false license plate and security vehicle detection and identification system finds that the electronic license plate does not match the security license plate during the detection process, it will send out an alarm message for the security department to trace.
The intelligent traffic management system based on license plate recognition can provide safe and detailed classified traffic statistics for urban road planning and design, realize the safety optimization design of road planning management, and reduce traffic congestion black holes. The intelligent traffic management system can realize the sampling of vehicle traffic data at major urban intersections, and analyze vehicle categories. Such as buses, trucks, buses, cars, taxis, etc. And traffic flow, and provide safety data such as traffic flow, vehicle type, peak period, and peak value for road planning and design, and scientifically guide road planning.
The intelligent traffic management system based on license plate recognition can better solve various “persistent” problems in current traffic management.
Does anyone know the intelligent license plate recognition system?
Summary of Mercedes-Benz technicians;
The intelligent license plate recognition system can make the entrance and exit management of the parking lot intelligent: install the license plate recognition equipment at the entrance and exit, record the license plate number and the time of entry and exit of the vehicle, and combine it with the intelligent access control equipment to realize the automatic management of the vehicle. It can realize automatic timing charging, and can also automatically calculate the number of available parking spaces and give prompts to realize automatic management of parking fees, save manpower and improve efficiency.
Automatic release saves manpower and material resources:
Input the specified license plate information into the system. Automatically read the license plates of passing vehicles, and query the internal database. For the vehicle system that requires an automatic clearance, drive the intelligent gate to let it pass. For other vehicle systems, the on-duty personnel will issue a warning and deal with it. It can be used in special units (such as military management areas, secret units, key protection units, etc.), road and bridge toll stations, high-end residential areas, etc.