The Application And Research Of Fuzzy Cognitive Map Optimization Algorithm And Geometry Recognition

At present, the Fuzzy Cognitive Map (FCM) learning algorithm and itsapplication research has been a great development. The learning algorithm of FCMis still an emerging field, has not yet formed an independent branch. The biggestdrawback of the FCM is based on prior knowledge, and the prior knowledge mayhave a great impact on reasoning results, and may converge to a potential undesiredsteady state. Elimination of FCM disadvantages, such as the abstract estimate of theinitial weight matrix and relies on the subjective reasoning of expert knowledge, willsignificantly improve its performance. Strengthen FCM learning to obtain the trustedweight of FCM, and enhance the adaptability and robustness of FCM model anddynamic simulation. Therefore, FCM learning is an important issue. In field ofcomputer vision, the mainstream recognition methods for basic geometry have twomain classes, such as region-based shape recognition, and based on the descriptionof object outline in the image. Both types of methods deal with issues under prioriknowledge, such as blocking and loss of the graphic. And the existing recognitionmethods have the disadvantages, such as large calculated amount and rigid matching.The graphic recognition method based on Fuzzy Cognitive Map (FCM) can avoid theproblems effectively. Now the study in basic geometric pattern recognition based onFCM has a long way to go for lacking further research.This article adopts Ant Colony Optimization (ACO) algorithm to discuss theoptimize research on FCM weight matrix, and ACO algorithm was used to solve theminimum by the defined object function to finish the study task of FCM. In theexperiment, we use a typical problem in process control to test and verify thelearning method for another purpose. The method overcomes some disadvantages onother FCM learning methods and rich the application of swarm intelligenceoptimization on FCM. We also compared the experiment results to the ParticleSwarm Optimization (PSO) algorithm on the optimized FCM weight matrix; theresults show that, this method has advantages on effectiveness and robustness.For a better perfection on FCM recognition model which designed aim at basicgeometric image. We optimized the existing weight matrix based on quadrilateraland triangle FCM recognition model. The experiment results show that it has highercognition rate after the optimized treatment on image deficiency and shading problems.