摘要
针对现行的图像分解方法对于分割精确的运动目标可以准确地进行分解,但当运动环境复杂、运动目标分割过程中受到阴影干扰而导致运动目标形状特征发生突变,进一步导致图像分解抗噪性能差,阴影区域去除不完全、分割精度低等问题,提出一种基于行为视觉的运动过程合理化图像分解方法。采用背景去除法提取运动区域,依据运动目标几何特征对是否存在运动区域阴影部分进行判断及粗分割,结合区域一致性测度方法对阴影区域中全影及半影进行有效检测和去除。依据运动目标外观表征构建了一种能量变化图,提取出描述运动目标形状信息和运动信息的行为特征,利用运动行为特征作为聚类中心,采用人工蜂群模糊聚类方法求解运动过程图像中的最优聚类中心,依据最大隶属度原则对运动过程图像进行分解。实验结果表明,该方法有效去除了运动目标图像阴影区域,具有抗噪性强、分解精度高等优点。
In allusion to the problem that the current image decomposition method can accurately decompose the moving targets with segmentation precision,but when the motion environment is complex and the segmentation process of moving targets is interfered by shadow,the shape characteristics of moving targets change suddenly,and the anti-noise performance of image decomposition becomes worse with incomplete shadow area removal and low segmentation precision,a motion process rationalization method for image decomposition based on behavioral vision is proposed. The background removal method is adopted to extract the motion area. The shadow area of the motion area is judged and coarsely divided according to the geometric features of moving targets. The full shadows and penumbras in the shadow area are effectively detected and removed with the region consistency measurement method. According to the appearance of moving targets,an energy variation map is constructed to extract the behavioral features that describe the shape information and motion information of moving targets. With motion behavior features as the clustering center,the artificial bee swarm fuzzy clustering method is used to solve the best cluster center in motion process images. According to the maximum membership degree principle,the motion process images are decomposed. The experimental results show that this method can effectively remove the shadow area of moving target images and has the advantages of strong anti-noise performance and high decomposition precision.
出处
《现代电子技术》
北大核心
2018年第4期180-182,186,共4页
Modern Electronics Technique
基金
国家青年科学基金(81102198)~~
关键词
行为视觉
运动过程
合理化
图像分解
分割精度
聚类中心
behavioral vision
motion process
rationalization
image decomposition
segmentation precision
clustering center