Improved Kernel Fuzzy Clustering Algorithm And Its Application In Meteorology Target Segmentation

For a long time, weather has been closely linked with people’s activities, disastrous weather will adversely affect people’s lives, the timely forecast of it can effectively prevent unnecessary losses in all aspects, therefore, the research and application of Doppler weather radar images or meteorological images are received much attention, it helps people monitoring and predicting the disastrous weather in time. Due to the particularity of meteorological images, traditional image segmentation methods will appear incomplete, long computing time and other problems in extracting the target, so they can not meet the actual needs. Fuzzy theory has been found as better problem-solving skill, and the process of image segmentation is also the process of pixel classification, in recent years, many scholars use clustering algorithm to segment images, it improves the accuracy as opposed to the traditional image segmentation methods, but fuzzy clustering segmentation algorithm is still insufficient.For the application of the meteorological images, based on the kernel fuzzy clustering algorithm, this paper studies its lack, the details are:First, the basic principle and development of kernel fuzzy clustering are introduced, and then analyzing and summarizing the advantages and disadvantages of kernel fuzzy clustering according to the experiment.Second, in order to solve the problem that KFCM algorithm will be affected by the initial value, the ant colony algorithm is added to calculate the clustering number and the initial centers that needed to be set by the clustering algorithm, then use it to KFCM algorithm, it overcomes the insufficient of the kernel fuzzy clustering and reduces the effect of the initial parameters to the segmentation. As the introduction of the ant colony algorithm, the computation efficiency of the algorithm is lower, than the genetic algorithm is added to the ant colony clustering algorithm, it according to introduce the path mutation operator, and use the new pheromone update strategy to improve the algorithm. This new algorithm can increase the clustering precision and reduce the calculating time.Third, for the kernel fuzzy clustering algorithm, the impact of the weighting factor m to the image segmentation result is studied according to a large number of literature and experiments, the result shows the algorithm achieves best effect when m is in the interval of [1,3]. Then more experiments are carried out to solve the problem of selecting suitable parameters in ant colony algorithm and the genetic algorithm.At last, the features of the meteorological images are analyzed, the improved algorithm for the meteorological image application is discussed and the algorithm is evaluated. According to analyze the experiment, it shows that the ameliorative arithmetic can better manipulate the multi-target images and it has accurate segmentation to meet the actual needs.