Review of Image Intelligent Recognition Technology

Abstract

In summary, as the computer improved, people need it to solve more complex problems. Thus, computer scientist invented artificial neural network. At first, ANN is very simple and only have single hidden layer. After decades development, ANN become more powerful and smart. With more algorithms have been invented and more hidden layers, Convolutional Neural Network have become a widely used nowadays: it could use for image classification, such as writings and photos. It could also use in autonomous vehicles. CNN could help doctor to diagnose disease and increase device security by facial recognition. It could also use on weather forecasting. One important usage of CNN is license plate recognition. There are four steps in this recognition: plate localization, plate inclination correction, character segmentation and character recognition. First, computers need to find vehicle plate in the photos since they are taken in various conditions. This requires a series of preprocessing. After locating plate in the photos, the computer would calculate the angle of inclination of the plate and correct it. Next, the computer would locate characters on the plate and size of them, using such information to segment them. Last, CNN analyzes these characters and gives out results. A CNN is consisted by input layer, hidden layer and output layer. CNN would adjust the weights in each layer based on difference between the output and the answer. More adjustment would lead to more accurate result. Finally, CNN would identify characters on plates. In conclusion, CNN has a bright prospect of application, and it will make things more convenient.