Extraction of Iris Curling Wheel Based on Two Stage ant Colony Algorithm.

In iridology, the iris curling wheel is an important diagnostic marker. It is located in the periphery of the intestinal ring and is a circular line. It divides the iris into two parts: the central pupil and the peripheral ciliary body. It is composed of blood vessels and nerves, representing the autonomic visceral nervous system in the human nervous system (sympathetic and parasympathetic), it can display a lot of information related to the intestinal and visceral nerves. When the edge of the roll wheel is abnormal, such as swelling or contraction, it indicates abnormal bowel function. So the health of the human  can be detected by detecting the change in the curling wheel. Therefore, accurately detecting the edge of the reeled wheel is the premise for further pathological analysis and detection. In addition, in the computer diagnosis system based on iris, the position of the curling wheel is the basis of realizing the iris map coverage. Therefore, whether the exact position of iris rollers can be extracted is also an important prerequisite to determine the performance of computer diagnosis system based on iris.

Ant colony algorithm (ACA) is a simulated evolutionary algorithm, which is called basic ant algorithm (AS),proposed by Italian scholar Dorigo M and others in 1991. The basic idea of ACA is that several ants start from the source (ant nest) to the target continuously to carry food, they sprinkle pheromones on the way they passed. These pheromones evaporate over time. The pheromone concentration on the path affects the direction of the subsequent ant's path. The greater the pheromone concentration on the path, the greater the probability of being selected by the ant. As a result, the shortest path from source to destination per unit time is repeated by the ant more times, and the relative pheromone concentration on the path is greater. After a period of time, the biggest path of pheromone is the optimal path from source point to target. At present, ant colony algorithm has been successfully applied to solve many combinatorial optimization problems, such as vehicle scheduling problem, traveling salesman problem, pipeline problem. In recent years, many scholars have applied AS algorithm to image processing because of its characteristics of positive feedback, distributed computing and heuristic search.

In view of the characteristics of the curling wheel, we propose a method based on the two stage ant colony algorithm to extract the curling wheel. The first stage uses a few iterations of more ants to determine the approximate initial position of the next stage. The second stage is to use fewer ants to extract the final contour of the curling wheel . The experimental results show that the algorithm can not only extract the curling wheel with clear edge, but also extract the crimping wheel with fuzzy edge and l with pit disturbance nearby.

College of Computer Science and Technology Zhu Lijun