摘要
为了提高变化光照下视觉特征的识别能力,提出一种光照自适应的特征提取方法。动态选择颜色、边缘特征建立特征集,解决单一类特征造成的跟踪不稳定性问题;定义局部光照变化测定函数,检测光照变化;利用Fisher准则评价特征识别能力,在线更新特征集,克服光强变化、噪声等因素的影响;采用模糊方法根据光强变化控制特征集的更新度。应用特征提取方法在粒子滤波框架下进行跟踪实验,在光照不断变化情况下能够实现稳定的目标跟踪,准确率较基于颜色特征的视觉跟踪方法有所提高。
To improve identification ability of visual features when illumination varies, the feature extracting method of self-adaptive illumination based on fuzzy control was proposed. Unstable track problem caused by using unique kind of features could be settled by building feature set through dynamically selecting color features and edge features. Local illumination varying detection function was determined to detect illumination variation. Fisher criterion was used to evaluate feature identification ability and renew feature set online. By doing so, the problems of illumination variation and noise factor could be settled. Fuzzy control method was used to control renewing rate of feature set according to illumination variation. Stable tracking target can be realized when illumination varies. Compared with color feature-based visual tracking method, its accuracy is increased.
出处
《计算机应用》
CSCD
北大核心
2010年第9期2438-2440,2457,共4页
journal of Computer Applications
基金
国家自然科学基金资助项目(60775058)
教育部科学技术研究重点项目(107028)
北京市教育委员会科技计划面上项目(KM201010772021)
