一种自适应图像数据源的花卉图像分割方法

A FLOWER IMAGES SEGMENTATION METHOD BASED ON ADAPTIVE IMAGE DATA SOURCES

为解决自然场景下各种颜色花朵目标的提取问题,提出一种基于自适应图像数据源的彩色图像分割方法。该方法中,一幅图像的不同数据源对应着此图像不同的色彩分量信息。由于模糊C均值聚类算法(FCM)是一种局部搜索算法,因此新方法中首先利用细菌觅食优化算法(BFO)的全局寻优性与FCM结合,从而寻得一幅图像在多个色彩分量下各自的全局最优聚类中心;然后利用一种新的综合的模糊聚类评价函数求得此图像在其不同色彩分量下即不同数据源下的分类质量;最后输出分类质量最好的分类结果。通过对真实场景中采集的10幅具有代表性的图像进行实验,结果证明新方案能适应目标颜色和背景因素的变化,分割出更接近期望的目标。

To solve the problem of extracting the goal of flowers in various colours under natural scene, this paper presents a colourful image segmentation method which is based on self-adaptive image data source. In this method, different data sources of an image correspond to different colour component information in image. Since fuzzy c-means clustering algorithm （FCM） is a local search algorithm, therefore the new method first uses the global optimisation property of bacterial foraging optimisation （BFO） algorithm and combines it with FCM ,so as to find the respective global optimal cluster centre of an image under multiple colour components; then, the new method uses a new integrated fuzzy clustering evaluation function to obtain the classification quality of the image under different colour components, i. e. different data sources; finally, the new method, outputs the classification results with best classification quality. The paper experiments on ten representative images in real scene acquisition,results prove that the new scheme can adapt to the changes of the target colours and background factors,and segments the targets closer to expected goal.