by:Shenzhen TGW Technology Co.,Ltd
With the rapid development of society and the economy, cars are playing an important role in people's work, production and life. The overall number of cars is increasing rapidly, transportation has become more and more difficult to be managed, traffic problems have also emerged. what is more, these problems are becoming More and more complicated.
With the emergence of Alpr software, roads, vehicles, and users were closely unified. Alpr software played an all-round role in society and was a real-time, accurate, and efficient car management system. The most important thing about Alpr software is the license plate recognition algorithm.
The identification algorithm is a technology with the highest gold content in the entire system, and its research and development costs also occupy the bulk of the entire product research and development costs. The research and development team spends the most effort and time on the algorithm.
With the development of science and technology, a variety of color photos and color technologies have become widespread, but these color images contain a lot of irrelevant information. For the recognition algorithm, the color image will cause a lot of interference, so the algorithm needs to gray it.
(1) Edge detection
Image edges are the basis for image segmentation, target area recognition, and area shape extraction. They also play a very important role in extracting license plate locations, so edge detection is very necessary.
(2) Image corrosion
Image erosion can erode the edge image. After the erosion, an image of the target area can be obtained, which greatly reduces the target range and improves the accuracy of the license plate recognition algorithm.
(3) Fill the image
Filling an image is a closed operation on the image, used to smooth the outline of the image, and used to fill tiny holes, cracks, broken targets, etc. in the target.
(4) Morphological filtering
Finally, morphological filtering is used to remove a large number of irrelevant small objects, and the rest are the targets that need to be identified.
After the pre-processed images, the rest are rectangular pictures of different sizes. You only need to find the rectangular figure of the license plate in these graphics to implement the next step. By some calculations, the position of the license plate can be found out, and the content of the rectangle can be extracted.
Calculate the sum of each column by column from left to right through the binary image, distinguish it by using 1 and 0, and finally split each character out.
After the characters are segmented, how does the algorithm need to recognize each character? First, establish a recognition sample database, read the segmented and normalized characters, match the cut characters with the template library, and finally get all the data of the segmented characters.
For more info,please click here to wikipedia.