动态多目标优化的运动物体图像分割

对小区背景下运动物体图像进行分割时多使用单目标或多目标优化方法,这类方法不能有效适应目标的动态变化,因此本文提出一种动态多目标图像分割优化方法。该方法将时间及环境动态因素作为动态因子,利用K均值(KMeans)算法和和模糊C均值(FCM)聚类算法构造多目标函数;结合动态多目标粒子群算法(DMPSO),使用背景差分法定义环境变化规则,实现动态多目标的图像分割。根据DMPSO算法优化后的聚类结果,分别与K-Means和FCM聚类方法得到的结果进行了对比。结果表明,动态多目标优化的Pareto最优解集分布均匀,图像分割准确率可达到95%,对图像识别的准确率可达到90%,具有较高的识别能力,能满足确定背景下运动物体的准确识别。

Moving object image segmentation by dynamic multi-objective optimization

The single objective and multi-objective optimization methods are usually adopted to segment the moving objects in community background images. However, these methods can not adapt to the dynamic change of the objects effectively. In this paper, a dynamic multi-objective optimization image segmentation method is proposed. The method makes use of the time and environment dynamic changes as dynamic factors, and takes the advantages of the K-Means and Fuzzy C-Means (FCM) clustering algorithms to construct the multi-objective function. In addition, the Dynamic Multi-objective Particle Swarm Optimization (DMPSO) algorithm is also embedded in the method, and background difference method is used to define environmental change rules to implement dynamic multi-objective image segmentation. The simulation results based on the DMPSO algorithm are compared with that of K-Means and FCM algorithms. The results show that the dynamic multi-objective optimization has made the Pareto optimal solution set evenly distributed as compared with single target segmentation algorithm, the accuracy of image segmentation reaches 95%, and the recognition accuracy reaches 90%. For the high recognition capability, the algorithm satisfies the accurate identification of moving objects under the determined background.