National unified customer service hotline:
18071134559 (Manager Yu) 15871828100 (Manager Zhang) 13986033813 (Manager Xu) 18617107500 (Manager Xu)
In recent years, automatic license plate recognition technology has been widely used in urban intelligent transportation systems, such as snapping red lights, capturing speeding violations, and traffic security bayonet systems, especially traffic security bayonet systems, which are important technologies for public security criminal investigation management. One of the methods, put forward higher requirements for the automatic recognition of license plates, and promote the rapid development of automatic recognition of license plates.
Although the New Year's Day holiday in 2016 did not achieve free high-speed passenger cars with less than 7 seats, some popular high-speed sections were still blocked as parking lots, but most of them improved overall. Many travellers claim that after ETC realized the nationwide network, using the ETC channel to pass through high-speed toll stations has reduced the interruption time by more than 10 seconds, which has also become a major reason for reducing high-speed congestion.
License Plate Recognition
The ETC channel enables fast cars to pass quickly. The importance of the automatic railing function is self-evident, but the automatic license plate recognition technology also plays an irreplaceable role. In recent years, automatic license plate recognition technology has been widely used in urban intelligent transportation systems, such as snapping red lights, capturing speeding violations, and traffic security bayonet systems, especially traffic security bayonet systems, which are important technologies for public security criminal investigation management. One of the methods, put forward higher requirements for the automatic recognition of license plates, and promote the rapid development of automatic recognition of license plates.
License plate recognition technology (LicensePlateRecognition, LPR) is based on computer technology, image processing technology, and fuzzy recognition. It establishes a vehicle characteristic model and identifies vehicle characteristics, such as license plates, vehicle models, and colors. It is a special computer vision system that takes a specific target as an object. It can automatically extract the license plate image from an image, automatically segment the characters, and then recognize the characters. It uses advanced image processing, pattern recognition and artificial intelligence technology. The collected image information is processed to automatically and accurately identify the numbers, letters, and Chinese characters of the license plate in real time, and directly give the recognition results, making computerized monitoring and management of vehicles a reality.
Compared with the traditional RF card swiping management system, the biggest advantages of the automatic license plate recognition system are: first, it can completely realize non-interference and non-parking traffic; second, it truly realizes the traffic management requirements of one vehicle and one vehicle to prevent transmission The occurrence of vehicle-card swaps; third, the realization of fully computerized records and statistics of parking lot management fees, minimizing the loss of parking fees.
Automatic license plate recognition process
The front end of the system uses an embedded high-definition integrated camera, which can achieve millions of resolutions of video and picture bitstream output, a built-in high-performance DSP chip, support for built-in intelligent algorithms, video detection, automatic license plate recognition and other functions.
The automatic license plate recognition system with a built-in camera uses a unique texture + model algorithm. It has the characteristics of accurate positioning, fast recognition speed, high recognition accuracy, and low misrecognition rate. Perform normal capture. Transplanting the license plate recognition algorithm based on the back-end server or front-end industrial computer in the traditional model to the front-end camera, it has the characteristics of high integration, high stability, high adaptability, etc. The complex environment of actual roads can better meet the requirements of all-weather work in intelligent transportation systems.
Dynamic video recognition technology is used to identify each frame of the video stream, thereby increasing the number of recognition comparisons and greatly improving the efficiency and accuracy of recognition.
The automatic recognition of vehicle license plates is mainly based on image segmentation and image recognition theory. An analysis is performed on the image containing the vehicle license plate to determine the position of the license plate in the image and further extract and recognize text characters.
The specific steps of recognition are divided into license plate location, license plate extraction, and character recognition. In the natural environment, the camera first performs a large-scale correlation search on the collected video images, finds several areas that match the characteristics of the vehicle license plate as candidate areas, then further analyzes and judges these candidate areas, and finally selects the best one. Use the area as a license plate area and segment it from the image.
After locating the license plate area, the license plate area is divided into individual characters and then recognized. The license plate recognition algorithm uses a template-based matching algorithm to first binarize the segmented characters and scale them into templates in the character database. Size, then match all templates, and finally select the best match as the result. Through this multiple comparison method, the accuracy of license plate recognition is greatly improved.
Factors Affecting Automatic License Plate Recognition
So what are the factors that affect the results of license plate recognition?
1.The impact of images on recognition
Only when the image is in sharp focus can the recognition result achieve a satisfactory result. When the license plate size in the image is 150 × 40 dot matrix, the characters are not easy to stick, which is helpful for segmentation, and the characters of the segmented characters are more obvious, which is helpful for recognition.
2.The effect of vehicle speed on the image
The PAL video standard adopted by our country stipulates 25 frames per second (the interval between each frame is 40 milliseconds). If the depth of field (clear image range) of the camera lens is 1.0 m. For a speed of 90 km / h (0.001 * 3600 / 0.04 = 90), the camera can only capture a clear picture of 1 frame. In order to get more clear images, you should try to increase the depth of field. The specific method is to increase the lens and reduce the full size of the CCD camera.
3, the impact of the trigger device on the image
In a Windows multitasking system, the task switching time is extremely unstable. After measurement, this time ranges from 0 to several hundred milliseconds. When the system CPU usage is high, the task switching time is very long. The CPU usage of video application systems is relatively high. The trigger device actually uses the task query method to detect the status of the I / O port, and when the set conditions are reached, the image is captured. Because the I / O port state has a certain delay from detection, the captured image may not be the clearest. This situation is more obvious when the vehicle speed is high.
4, the impact of models on the image
The influence of the vehicle model is obvious regardless of whether the ground-sensing coil or infrared trigger is used. The trigger position of the big car and the small car may exceed 0.5 meters even at low speed. When it exceeds 1 meter, the captured image may not be the clearest. of.
5.Why is the recognition speed fast?
On urban roads and national roads, the speed is generally 0 to 80 km / h. The actual vehicle detection application system should reach this speed level, and the speed should not be limited. At a speed of 80 km / h, the camera has only one clear image. The only way to accurately capture this image is to capture and identify the image frame by frame. This requires less time to capture and identify. In 40 milliseconds, in order to ensure the normal operation of the Windows system, this time had to be shortened to less than 20 milliseconds.
Application characteristics of automatic license plate recognition system
1.Strong ISP processing capability
The recognition rate of the automatic license plate recognition system has a close relationship with the quality of the license plate and the quality of the images taken. Not only do various factors such as rust, dirt, paint peeling, font fading, etc., greatly affect the accuracy of license plate recognition, but Whether the shooting environment is ideal will also have a great impact on license plate recognition.
The intelligent traffic camera has a powerful built-in ISP processing function, which can provide functions such as video stabilization, face detection, noise filtering, automatic white balance, automatic exposure, and gamma correction, edge enhancement, etc., bringing the image quality and effects to a new level. Not only improves the user's actual perception, but also provides a good calculation and analysis basis for more intelligent applications such as license plate recognition, which fully guarantees a high accuracy of license plate recognition.
2.Good adaptability to light climate background
Many license plate recognition systems have a higher recognition rate on cloudy days, but fall or even fail to recognize on sunny days. In the case of direct light, the shooting direction is the same as the direction of sunlight. The license plate area is very bright, resulting in thicker strokes and sticking to each other. Moreover, the license plates of our country use reflective paint. In severe cases, mirror reflection may occur, and the license plate number cannot be seen clearly . In addition, bright lines and halos produced by reflections on the surface of the car body will also affect recognition. The license plate recognition is mostly used to identify a moving vehicle. The license plate area is not fixed in the entire image. Ordinary cameras cannot adjust the license plate area. At night, the vehicle turns on the lights. The ordinary camera is affected by the headlights and weakens the exposure intensity. As a result, the image license plate area is very dark and the number cannot be seen. The light from the car headlights may also form a large area to block the license plate area.
After years of development, the automatic license plate recognition system has become a relatively mature technology. The traditional license plate recognition system is based on simulating standard definition images for detection and recognition. Due to the low resolution of SD images, the lack of layering, and the small field of view, the license plate recognition cannot achieve the desired effect. Often, in order to achieve the license plate recognition rate, The vehicle panorama needs to be sacrificed, so two cameras need to be used to complete the close-up of the license plate and record the vehicle panorama. The system complexity is high.
It is believed that in the next few years, with the continuous application and construction of high-definition intelligent transportation systems in various places, the automatic license plate recognition technology will gradually develop toward high-definition, integration, and intelligence. In each application system, it will continue to play its increasingly important role. Role.
Copyright © 2019-2020 Wuhan Chinasoft Intelligent Technology Co., Ltd.
Headquarters Address: 15th Floor, Wuhan Optics Valley United International Building Contact: 027-87568430
Factory Address: Building A8, Xinbosheng (Jinmi) Industrial Park, No. 179 South Shahu Avenue, Tangxia Town, Dongguan City
Sales manager: 18071134559 (manager Yu) 15871828100 (manager Zhang)
13986033813 (Manager Xu) 18617107500 (Manager Xu)Record number: Hubei ICP 07500413-11