Based On Ant Colony Algorithm For Image Edge Detection

Edge points are one the most basic features of the image, it contains much useful information for the image identification, and its provides a valuable and important information for people to describe and explain or identify the images. Edge detection is a classic research topic in computer vision and image processing, the aim is to find the shape and the reflection or transmission information, it is the basic step in the image processing, image analysis, pattern recognition and computer vision areas. The correctness and reliability directly affect the machine vision system for objective understanding of the world. The target edge point, texture and even noise may become a significant edge point, so it is difficult to find a universal edge detection algorithm. Many existing methods have their own characteristics, but also have their limitations and deficiencies, so this field remains to be further improvement and development.Traditional edge detection algorithms based on ant colony algorithm are easily fallen into the local optimal solutions, and have inaccurate edge location. There are some reasons, firstly, noise and edge points are all in the gray mutations, so that the ants choice them with the same probability, fail to effectively suppress the noise. Secondly, the ants choose random initial position, making a lot of ants distribution within the background in independent calculation, reducing the efficiency of the algorithm. Thirdly, a huge number of calculations are needed. Overall, the fundamental reason that causes many defects of the ant colony algorithm is the imbalance between the positive feedback and the randomness.In order to reduce the calculation of the traditional ant colony algorithm, analyzed the traditional edge detection algorithms based on differential: 1) the edge location is not accurate; 2) some diagonal points are undetected. In order to overcome this short-comings, analyzed the formation mechanism of the Hilbert transform, proposed using it to extract diagonal points. First of all ,used the Canny edge detection algorithm to extract the edge points , then got the threshold through a linear combination of the high and the low threshold, to get the prior knowledge by the new threshold. Second, extracted diagonal points through the Hilbert transform. Third, let the edge points and the diagonal points as heuristic information, established the edge tracking model. Finally, get single pixel edges through some related mechanisms. This model achieves the guiding function via the pheromone and the heuristic information, avoids the ants trapped in local optimal solutions. Experimental results show that this algorithm has better accuracy, strong noise robustness and fast positioning speed.