Copyright © Shenzhen Qianhai Intellidata Technology Co., Ltd. All Rights Reserved. Site Map
Abstract: With the rapid growth of urban population, the rapid increase in the number of vehicles owned by people, and the implementation of government regulations, the license plate recognition market has a strong growth momentum. The parking management industry not only continues to develop in the direction of high-end, unmanned and intelligent, but also full-video, fast-pass, and unattended parking management systems are accelerating on the market.
[CPS Zhongan Network cps.com.cn] With the rapid growth of urban population, the rapid increase in the number of vehicles owned by people, and the implementation of government regulations, the license plate recognition market has a strong growth momentum. The parking management industry not only continues to develop in the direction of high-end, unmanned and intelligent, but also full-video, fast-pass, and unattended parking management systems are accelerating on the market.
The premise and key to the intelligent process of parking management is the license plate recognition camera. Among the license plate camera manufacturers, Intellidata can be said to be one of the best. Because Intellidata is a rare scientific and technological innovation enterprise in the industry that integrates the research and development, production, and sales of intelligent products. It leads the industry development trend in terms of technology, quality, function, innovation and other aspects. It is also the first domestic manufacturer to use domestic chips to make license plate recognition cameras with independent algorithms.
At present, Intellidata has a variety of license plate recognition camera products. VW83-C is a camera tailored for cloud billing solutions. It has a fully open intelligent hardware platform, rich interfaces, and simple debugging. It is equipped with a wireless transmission 4G module, which provides remote debugging and cloud operation and maintenance functions for the license plate recognition system. Not long ago, we were fortunate to obtain this new product and tested it:
[Occasions and Equipment]
Scenes: roads, gates, basements.
Equipment: UPS, batteries, light meters, etc.
[Unboxing]
Gif1
VW83-C uses a fixed-focus lens, which can support a recognition distance of 2.8 meters to 10 meters without focusing. The system will automatically adjust the algorithm recognition area and algorithm recognition related parameters according to the recognition distance, making it easier for engineers to debug.
VW83-C supports SD card expansion and is connected to the industry's mainstream screen display and voice control card. It can realize offline screen display voice broadcast and contains offline billing function, which increases the reliability of project use.
[Street test]
The street scene is mainly used to test the continuous capture performance, so the sensitivity is set to the highest and the capture interval is set to the minimum.
In the case of sufficient light during the day, VW83-C can achieve continuous license plate capture without pressure.
https://v.qq.com/x/page/i0851xogs2g.html?
In the night scene, due to insufficient light, the capture performance may be affected. Since VW83-C is designed to be used mainly in parking lots and gates, in order to obtain the actual performance of VW83-C in low-light scenes, the test scene was transferred to the basement of the community:
[Basement Test]
The low-light test scene was selected in the basement, and the light intensity was about 0.22Lux. The low-light test is divided into two groups. In the first group of tests, the vehicle's high beam is turned on, and the capture situation is tested through the camera monitoring area, in the second group of tests, the vehicle's high beam is turned off, only the daytime running lights are left on, and the camera monitoring area is passed and tested. (The red frame shows the real-time scene, and the video is slowed down by 0.3 times when passing through)
https://v.qq.com/x/page/d08544kf3x4.html?
In the basement scene, the high beam is the only light source that illuminates the camera lens, however, it can be seen in the video that when the high beam is turned on, the camera's function and speed are not affected when facing strong light and backlight capture. It can be seen from the WEB page that the license plate is captured and recognized as soon as it enters the lens recognition area.
After turning off the headlights, the light intensity in the basement environment is very dark. At this time, only the vehicle's daytime running lights remain on, which tests the performance of the camera under low illumination. It can be seen from the video that the speed of the camera capturing the license plate has decreased compared with the first group of tests, but it still achieves automatic adjustment of the ISP for recognition according to different scenes.
[Barrier]
In the barrier test, the camera's large-angle capture capability is mainly observed. Under normal circumstances, the width value of the license plate image is 200 pixels, if the angle is greater than 60°, the width value of the license plate image will be less than 100 pixels, and the resolution will be greatly reduced. In addition, the license plate may have horizontal tilt, vertical tilt and trapezoidal distortion. In this case, the license plate image is often not rectangular, which will seriously affect the accuracy of license plate positioning and character segmentation. Therefore, the problem of license plate imaging deformation recognition caused by large angles has become one of the difficulties recognized by the industry.
During the test, the position of the tripod is fixed. By continuously adjusting the tripod heading axis and recording the scale and recognition results, the approximate license plate capture limit angle can be obtained:
It should be noted that the test here uses "video trigger". Although "video trigger" is convenient and simple, it is slightly inferior in trigger rate and recognition efficiency due to the extreme nature of the algorithm. Therefore, most license plate manufacturers use "dual cameras" or "external trigger" to ensure the stability and recognition efficiency of the camera in various extreme environments, and Intellidata uses its own algorithm to freeze the extreme capture angle at 73°, which is excellent.
[LAN/4G communication]
VW83-C has a USB interface reserved inside. After inserting the WiFi module to start the camera, you can use a dedicated APP to connect and control the license plate camera through the WiFi signal sent by the camera, which is similar to the WiFi connection function of consumer cameras, which is very convenient.
The settings that can be made include IP parameters, scene parameters, identification areas, etc.: Although it is not as detailed as the browser interface connected by network cable, it is convenient enough for basic settings. By building a WiFi connection, business data can be transmitted wirelessly, meeting the need for free wiring in parking lots.
[Identification and Anti-counterfeiting]
https://v.qq.com/x/page/m0854daj4z4.html?
At present, the face recognition technology in high-end projects is equipped with a liveness detection function so that it can deal with deception and attacks in the form of pictures and videos. For intelligent parking scenarios, it is very important to develop a set of effective license plate anti-counterfeiting technology to address the problems of using other people's license plate information to deceive the gate, or using fake license plate pictures to deceive license plate cameras, falsifying parking time, affecting billing, etc. Intellidata has prepared its own innovative Anti-Fake Technology license plate recognition and anti-counterfeiting technology for this purpose.
As can be seen from the video, the license plate camera with Anti-Fake Technology turned on no longer recognizes fake license plate pictures, and the effect is stable. The addition of this technology can increase the crackdown on fraudulent behaviors of forging, altering, and using fake license plates in picture styles, and provide strong support for effective order management and authenticity identification of vehicles and communities. It should be pointed out that if you need to use the Camera Assistant App (VW83-C mobile client) for unattended cloud billing/monitoring operations, you need to start the camera's Anti-Fake Technology through the WEB page first. The mobile client cannot be controlled directly and is not enabled by default.
[Summary]
As a cost-effective import and export product, VW83-C solves the major problem in complex scenes - "large-angle recognition", and also has many other highlights:
· Self-developed core algorithm, stable, high recognition rate, fast speed, the comprehensive recognition rate throughout the day reaches 99.8%, the recognition types exceed 1500+, and the recognition speed is <100ms.
· Complete range of license plate recognition, compatible with new energy and new embassy license plates, in line with the latest national GA36 motor vehicle license plate standard.
· IVE solidified chip, low power consumption, low heat generation, and camera is not easy to fog.
· Simple debugging and configuration, arbitrary scene switching, and automatic parameter configuration of the camera.
· The addition of cloud management, mobile debugging and other functions has injected new impetus into unattended parking.
[Appendix]
VW83-CV303 product specifications:
[License plate recognition principle]
1. Extract the license plate area of the captured image according to the license plate detection algorithm.
2. According to the region segmentation algorithm, separate the Chinese characters or character regions.
3. Use the convolutional neural network model trained with a large number of special license plate sample images to classify and recognize the separate Chinese characters and character regions.
4. Finally, output the recognition results.
Gif2
[Identification and Anti-Counterfeiting Principle]
Anti-Fake Technology identification and anti-counterfeiting technology mainly divides the following steps into identifying the validity of license plate information:
1. When a vehicle enters the video area, the entrance and exit cameras can intelligently collect the movement trajectory of the vehicle (license plate), and at the same time combine the license plate imaging pixel changes, use the cycle depth model for analysis, and determine whether it is a real vehicle.
2. While identifying and judging the movement trajectory of the license plate, the entrance and exit cameras can combine the vehicle model, use the CNN convolutional neural network algorithm to perform vehicle detection, and match whether it is a fake license plate deception.