PCB Infrared Image Segment Based On MRF And The Game-theory

The infrared image of PCB has many characteristics, such as large noise, heating zone of chips are cross.most of the existing image segmentation algorithm are aimed at visible light images or medical images, and specific algorithm deal to infrared images of PCB is rare.so traditional segmentation algorithm needs to be improved and optimized and improved in order to complete the PCB infrared image segment.First the paper describes the characteristics of the PCB infrared image, and introduces a variety of image segmentation algorithm. The key point of the segmentation of the infrared images also been discussed; then we introduced the traditional Markov model of segmentation algorithm based on this model. And we studied the segmentation algorithm based on the model. through simulation experiments, we analyzes the advantages and disadvantages of traditional methods and inadequate in dealing with the infrared image.In this paper, we brought the concept of game theory and algorithms based on Markov random MMD field together, proposed MRF segmentation method based on game theory.Evolved the process of segmentation as the game between pixels Reach the best optimal solution by finding Nash equilibrium point.Experiment shows: The new algorithm can complete segmentation of the PCB infrared image, dividing up the core area of the chip,and edge details of the heating area is also very clear. and reflects the heating zone and heating position of the chip, the result is much better than the traditional method of Gibbs sampling and ICM algorithm.In this paper,combining the game theory with traditional ICM, SA algorithm Proposed the segmentation algorithms based on the game of the two methods Via the game between the two Optimized strategy, forming a strong complementary segmentation module, make up the problem of ICM algorithm is easy to fall into local optimal. And the time consumption problem of the SA algorithm. The simulation proved that the new algorithm for solving the problem of infrared image segmentation is effective.