Iris Recognition Based on Artificial Intelligence

  Iris recognition system is composed of hardware and software. Hardware module provides iris image acquisition device and software algorithm platform; software module is further divided into preprocessing module, feature extraction and recognition module. The details are as follows:

  1. Hardware module: Iris acquisition device usually consists of four parts: optical lens, image sensor, light source and image transmission module. It provides data storage and operation, algorithm operation and other functions for software module. Therefore, when choosing hardware platform, we must consider the characteristics of iris recognition itself, such as frequent floating-point operations and large amount of data, and choose a suitable and effective platform.

  2. Software module: The preprocessing module usually consists of three processes: the first step is iris localization. Simply speaking, iris localization is to use the edge detection algorithm in digital image processing to obtain the range of the iris region in the image, and realize the separation of iris, pupil and sclera. The second step is normalization. The iris region is normalized so that the iris images collected under different conditions are mapped to the same scale space. The third step is illumination compensation. Relevant illumination compensation algorithm is used to process the normalized iris image to eliminate the influence of illumination. Feature extraction and recognition module feature extraction is the process of extracting limited features from normalized images to uniquely represent iris images. Feature extraction can uniquely represent different samples in either spatial or frequency domain. After determining the encoding features, the feature vectors are mapped by using the classification method corresponding to the encoding strategy, and the final recognition results are obtained in the pattern space where the feature vectors are located.

  Our laboratory has made many achievements in the research and development of iris acquisition equipment, the technological innovation of iris recognition system, and the theoretical research of iris recognition. Successfully developed a number of iris acquisition equipment and iris recognition system with independent intellectual property rights, which can be roughly divided into the following three categories: 1) small desktop iris acquisition instrument for popular science, teaching demonstration of the whole process of iris recognition. 2) Embedded iris recognition device for POS terminal identity authentication. 3) Hand-held iris collector, used in customs, airports, stations and other places where the flow of personnel is large and the safety requirements are high. At present, the iris recognition system composed of the above devices has been successfully applied to the daily attendance of laboratory students in the Computer College of Jilin University. The correct recognition rate of the system is 99.97%, which has reached the recognition accuracy of commercial iris recognition system. The second public iris bank in China has been established.

  At the same time, there are many theoretical achievements in iris preprocessing, iris feature expression and iris matching. It includes 1) quality evaluation of adaptive parameter selection based on local features 2) multi-scene iris segmentation method based on deep learning 3) feature coding under parameter adaptive Gabor filter 4) iris feature representation under multi-resolution 6) iris multi-feature fusion strategy based on score fusion.