What is the algorithm of license plate recognition system? Really?

Automatic recognition technology of automobile license plate

It is a pattern recognition technology that uses dynamic video or static images of vehicles to automatically recognize license plate numbers and license plate colors. Through the collection and processing of images, the automatic license plate recognition function is completed, and the license plate image can be automatically extracted from an image, the characters can be automatically divided, and then the characters can be recognized. Its hardware foundation generally includes trigger equipment (monitoring whether the vehicle enters the field of vision), camera equipment, lighting equipment, image acquisition equipment, and a processor (such as a computer) that recognizes the license plate number.

Automatic identification technology is divided into hard identification and soft identification (in fact, the two are complementary)

"Hardware recognition" is a series of character processing on the collected pictures through independent hardware devices. At present, there are two kinds of hardware recognition in the parking lot system industry, namely, independent license plate recognizer and front-end hardware recognition. The front-end hardware recognition integrated camera embeds the traditional license plate recognizer into the camera, which realizes the integration of the front-end hardware and the camera, and perfectly realizes a series of tasks such as image acquisition, video streaming transmission, character recognition, road gate lifting and so on.

"Software recognition" can be understood as the recognition of license plate number by software, and the captured pictures can be recognized by installing supporting license plate recognition software on the computer. Its working mode is to take multiple photos continuously through the camera, select a clearer one, and then carry out character processing through computer software to realize license plate recognition. Because multiple photos need to be taken for each recognition, the speed of soft recognition is slow. Moreover, the soft recognition system requires extremely high pictures and must be extremely clear to achieve the desired effect. The system requires extremely high on-site environment and debugging quality, which is not suitable for many occasions with poor environment, and it is also very important to identify the placement position of equipment.

Comparison between soft recognition and hard recognition;

1, analyze the recognition pattern.

Hard identification system: video stream analysis and identification, real-time analysis of video streams within the monitoring range throughout the day;

Soft recognition system: image analysis and recognition, taking photos of vehicles that arrive within the specified range, and then analyzing the photos; When the vehicle position is not good, it is easy to make mistakes in identification.

2. Intelligent algorithm model

Hard recognition system: adopting intelligent fuzzy lattice recognition algorithm, the accuracy is higher, and the recognition rate is greater than 99.70%. Almost no human intervention is needed.

Soft recognition system: OCR/ font topology recognition algorithm, there will be frequent false recognition, the accuracy is less than 90%. You need to manually enter the correct number plate.

3. Reliability and stability:

Hard recognition system: the special recognizer adopts high-speed DSP of TI company and is controlled by dual CPU to ensure the reliability and stability of the system.

Software identification system: software identification is prone to frequent crashes and so on, which often requires restarting the computer, resulting in intermittent paralysis of the system.